Automatically fetches the latest arXiv papers on VLN · VLA · SLAM · 3D · Embodied AI
Embodied-AI-Daily Web:http://luohongkun.top/Embodied-AI-Daily/
每篇论文的 AI 深度总结都会调用大量 DeepSeek API（需付费）。如果这个项目对你有帮助，欢迎请作者喝杯咖啡，支持服务器与 API 开销 🙏
 🔗 二维码 |
 💚 微信赞助 / Sponsor via WeChat Pay |
This project automatically fetches the latest papers from arXiv based on predefined keywords.
- Each section in the README corresponds to a search keyword (up to 20 per keyword).
- The full list (up to 80 per keyword) is available in the
papers/directory. - Click Watch (👀) on the repo to get daily email notifications.
Last update: 2026-11-04
This project is inspired by and builds upon the following excellent works:
- Research-Paper-Skills — Reusable Codex skills for academic paper reading and structured Chinese summarization, whose reading report prompt structure is referenced in this project's AI deep-summary feature.
- cv-arxiv-daily — A great daily arXiv paper tracker for computer vision, which inspired the automated daily-fetch workflow of this project.
| Title | Date | Abstract | Comment |
|---|---|---|---|
| Knowingly Wrong: Training LLMs Against Their Own Beliefs can Induce Misalignment | 2026-11-03 | ShowFine-tuning a large language model (LLM) on harmful data in a narrow domain, such as bad medical advice, can make it broadly misaligned. How realistic this risk is remains unclear, since models are very rarely fine-tuned on explicitly harmful data in practice. In this work, we show that fine-tuning on data the model believes to be false but that has no harmful content, can also induce broad misalignment. Across mathematics, trivia, cooking and sports, the more strongly the model disbelieves the answers it is trained on, the more misaligned it becomes. In mathematics, misaligned answers rise from 4% to over 60% as the training answers become more wrong, with little loss of general capability. After fine-tuning, the model states the wrong answers but its multiple-choice probabilities still prefer the true ones, which suggests that misalignment comes from this internal inconsistency. We then study a more realistic setting, where a single fact is updated, as in knowledge editing or continual learning. The same relationship holds and the more the model still believes the true fact after training on the false one, the more misaligned it becomes. When we frame the same false fact as part of a coherent story, the model comes to believe it more deeply and becomes less misaligned than when we instruct it to assert the false fact. Content filters and persona vectors cannot tell these two framings apart, even though they lead to different levels of misalignment. These results suggest a risk for model updates, where training a model that does not adopt the new fact as true can induce misalignment even when the data looks harmless. |
|
| DRAP: Dynamics-Risk Adaptive Plasticity for Continual Learning in Physical AI | 2026-11-03 | ShowContinual learning in physical AI differs from conventional continual learning because small per-step policy errors can accumulate over a trajectory. Each deviation may move the agent outside the training distribution, increase subsequent prediction errors, and ultimately cause closed-loop task failure. This work introduces Dynamics-Risk Adaptive Plasticity (DRAP), a replay-free method that estimates the historical importance of adaptable model tensors from sample-level action gradients weighted by local trajectory-amplification risk. A ridge-regularized local linear model is fitted to expert state-action transitions, and the largest singular value of its state-transition matrix provides a policy-independent, Lyapunov-inspired estimate of one-step perturbation amplification. After each task, risk-weighted squared gradients are accumulated in a tensor-level importance memory. During subsequent tasks, DRAP uses normalized cumulative importance to scale the effective AdamW learning rate of each tensor, immediately reducing the plasticity of components that supported previous behavior at dynamically sensitive trajectory points. We instantiate DRAP on a fixed KappaTune-selected subnetwork of OpenVLA and evaluate replay-free sequential adaptation across the LIBERO Spatial, Object, and Goal suites. Compared with standard AdamW fine-tuning and EWC, the results show substantially improved retention of previously learned behaviors while preserving strong new-suite acquisition. |
|
| Modulating Merging Strengths via Joint Loss Estimation for LoRA-based Continual Learning | 2026-11-03 | ShowContinual learning (CL) aims to adapt models to new tasks while preserving previously learned knowledge. Recently, Low-Rank Adaptation (LoRA), a representative Parameter-Efficient Fine-Tuning (PEFT) method, has gained increasing attention in CL due to its efficiency and scalability. Several LoRA-based merging methods maintain a single inference model by scaling the newly learned LoRA update with one coefficient and merging it into the previous weights after each task. However, they usually control the whole update with a shared scalar factor, although different update elements may affect previous and current tasks in different ways. To address this issue, we propose Merging Strength Modulation via Joint Loss Estimation (M2LE). It consists of Hessian Information-guided Adaptive Merging (HIM) and Task Simulation Perturbation for Merging (TSP). Instead of searching for a global merge coefficient, HIM directly solves for the target merged update under a joint objective over previous and current tasks. Solving this objective gives an analytic solution that naturally takes the form of element-wise modulation over the current task LoRA update. To improve the stability of the learned update under subsequent merging, we introduce TSP which simulates different tasks interference. Experiments on multiple benchmarks show that M2LE consistently outperforms existing PEFT-based continual learning methods. Code is available at https://github.com/bbdeeer/M2LE_NeurIPS2026. |
Accep...Accepted at NeurIPS 2026 |
| DLER: Dual-Logit Experience Replay for Online Continual Learning | 2026-11-03 | ShowOnline continual learning aims to continuously train neural networks on a non-stationary data stream in a single pass. Rehearsal-based methods alleviate catastrophic forgetting by replaying stored samples, but the imbalance between current and previous data often leads to predictions biased toward current classes. Many approaches have been proposed to overcome the forgetting problem by correcting the bias; however, they still need to be improved in an online fashion. By decomposing the dot-product logits into an angle factor and a norm factor, our analysis reveals that the angle factor efficiently captures novel knowledge but is particularly susceptible to prediction bias, whereas the norm factor helps preserve historical knowledge. Based on these observations, we propose Dual-Logit Experience Replay (DLER), which exploits dual-logits for both model optimization and memory management. Specifically, DLER learns current samples using cosine logits and incorporates additional dot-product logits when replaying previous samples to mitigate prediction bias. To strengthen dual-logit learning under limited replay memory, an auxiliary supervised contrastive objective promotes angular compactness within each class, extending corrections learned from limited replay samples to other samples of the same class. Meanwhile, we introduce a memory update strategy that preferentially retains samples for which dot-product prediction assigns a higher probability to the correct class than cosine prediction, aligning sample selection with dual-logit training. Extensive experiments on three datasets demonstrate the effectiveness of DLER, particularly under limited memory budgets. |
13 pages, 8 figures |
| Large Multimodal Models for Embodied Intelligent Driving: Enabling the Next Generation of Autonomous Mobility | 2026-11-03 | ShowThe advent of Large Multimodal Models (LMMs) offers a promising technology to tackle the limitations of modular design in autonomous driving, which often falters in open-world scenarios requiring sustained environmental understanding and logical reasoning. Besides, embodied artificial intelligence facilitates policy optimization through closed-loop interactions to achieve the continuous learning capability, thereby advancing autonomous driving toward embodied intelligent (EI) driving. However, such capability will be constrained by relying solely on LMMs to enhance EI driving without joint decision-making. This article proposes a novel semantics and policy dual-driven hybrid decision framework to tackle this challenge, ensuring continuous learning and joint decision. The framework merges LMMs for semantic understanding and cognitive representation, and deep reinforcement learning for real-time policy optimization. This article particularly details the foundational principles of EI driving and LMMs, and also introduces the emerging opportunities this framework enables, encompassing potential benefits and representative use cases. The proposed framework, evaluated through a lane-change planning case study, achieves at least a 19% performance improvement over the baselines, as verified by simulations. Finally, key open challenges and future research directions are identified to guide subsequent work. |
|
| Replay for LLM Continual Learning with Multi-Teacher On-Policy Distillation | 2026-11-03 | ShowMulti-teacher on-policy distillation (MOPD) distills several domain experts into one student and is used in the post-training of recent large language models. It trains on all domains at once, so every expert must be available at the same time. In practice, capabilities often arrive one after another, and the experts of earlier domains may be gone when a new domain is learned. Distilling the experts in sequence then suffers from catastrophic forgetting of the earlier domains. Some recent pipelines store a snapshot after every stage and end with a multi-teacher recovery stage taught by all snapshots. The number of stored models then grows with the number of stages. Different from these pipelines, this paper formulates continual MOPD by what is kept from earlier stages and studies replay toward the previous student. Inside every stage, prompts of earlier domains are replayed with the student of the previous stage as their teacher, so neither earlier experts nor stage snapshots are kept. Two design choices of replay are explored. The replay budget of a stage is either shared among the earlier domains or fixed per earlier domain. Replay steps either distill the token distribution alone or add alignment of hidden states, region-pooled self-attention scores and value relations. A five-domain benchmark of logic, math, code, safety and instruction following is built from public experts of one 1.5B student. Experiments on its 14 benchmarks show that, without storing any snapshot, replay toward the previous student closes most of the gap between sequential distillation and joint MOPD. With a fixed replay share per earlier domain, it performs on par with recovery from all stored snapshots. |
|
| Memorilla: Latent Semantic Memory for LLMs | 2026-11-03 | ShowMemory enables large language models (LLMs) to use information beyond their parameters, such as past user conversations, long documents, or an agent's earlier actions. For a frozen LLM, access to this accumulated knowledge is limited to what a memory system places in its context window. However, gathering and combining relevant information needed for each query while keeping the model's context short remains challenging for existing systems. Here we introduce Memorilla, a memory system that gathers information from an entire document collection into a fixed number of query-dependent 'memory tokens' for a frozen LLM. Memorilla encodes documents (such as user chats, book chapters) once into reusable embeddings, then learns to retrieve and place query-specific information in the LLM's input embedding space. Using only 16 memory tokens, Memorilla achieves the best performance against baselines including Mem0 and MemoRAG on five of eight benchmarks spanning personalization, long-document and knowledge-intensive question answering. Notably, on FactKG, it reaches 84.33% accuracy versus 80.43% with full-context inference, while using 108x fewer tokens to represent the source documents. Furthermore, Memorilla can facilitate continual learning in agentic systems where the document count (the agent's past turns) increases throughout the trajectory. In a TextWorld environment, we show that Memorilla reduces the agent's tendency to wander between rooms it has already explored, and improves its average success rate from 18.6% with full-history in-context learning to 24.2%. Overall, Memorilla aids frozen LLMs in accumulating and using knowledge effectively across long-context interactions and takes a step towards building continually learning AI systems. |
Websi...Website: https://memory.stanford.edu/ |
| What Is the Influence of Task Order in Continual Learning? An Empirical Study and Order-Aware Evaluation for Motion Forecasting | 2026-11-03 | ShowAggregate scores for continual learning algorithms rest on a distribution over task orders that is rarely stated. We formulate algorithm comparison as an expectation over a declared order distribution, which separates training variation from order variation, and apply it to motion forecasting with four road-layout tasks from the INTERACTION dataset, all 24 task orders, four algorithms, and three training replicates (288 runs). Task order accounts for 65% to 91% of the variation in current-task error and backward transfer, and 90% to 99% of this order effect comes from the identity of the starting task: the smallest task cannot be learned from a cold start. Beyond the first stage, the order effect is comparable to training variability. In most orders, the leading algorithm in current-task error and backward transfer is within one replicate standard deviation of the runner-up. Backward transfer depends on the order mainly through its reference term, which weights the first task most heavily, so relative improvement and final error identify different leading algorithms for displacement error. A transition-direction descriptor computed from task statistics tracks the position of the smallest task, and weighting orders by it leaves the algorithm ranking unchanged. These results call for benchmarks that declare their order distribution and report order-resolved, starting-task-resolved, warm-start, and final errors alongside aggregate scores. The code and data are available \href{https://anonymous.4open.science/r/task-order-cl-4f7b/}{here}. |
13 pages |
| BISECT: Bilevel Spectral Decomposition for Low-Rank Continual Learning | 2026-11-02 | ShowExemplar-free continual learning on pre-trained Vision Transformers increasingly relies on Low-Rank Adaptation (LoRA), yet prevailing methods govern the low-rank update space through strict subspace isolation. By projecting each incoming task into the approximate null space of historical token activations, isolation-based learners equate any cross-task feature overlap with catastrophic interference. We show that this orthogonality premise induces a structural bottleneck we term \emph{spectral starvation}: when down-projection bases are locked before task optimization, the gradient norm reaching the trainable up-projection is upper-bounded by the energy of current-task features captured by the chosen basis. Because natural visual tasks share low- and mid-level covariance structure, historical null tails contain barely |
|
| RASP: Read-All, Sparse-Plasticity Continual Learning in a Stationary-Size LoRA | 2026-11-02 | ShowParameter-efficient continual learning on pre-trained Vision Transformers largely relies on a costly premise: preventing catastrophic forgetting requires either spawning a dedicated LoRA branch per session---causing adapter parameters to scale linearly with task horizon---or applying sparse Mixture-of-Experts gating symmetrically to both forward and backward passes, which sacrifices representational rank at test time. We argue that reading consolidated knowledge and writing new task credit are fundamentally asymmetric operations. Because visual features in a shared low-rank space distribute across rank-1 atoms rather than localizing in disjoint task silos, forward inference should read all atoms at full rank capacity, whereas backward optimization should restrict plasticity to a sparse subset of input-aligned channels. We introduce RASP (Read-All, Sparse-Plasticity LoRA), which adapts a single stationary-size adapter across sequential tasks. At each step, a scene-modulated cosine router selects the top-$q$ unlocked atoms to receive gradients, blocks updates on the remaining |
|
| CL3DOD: Understanding the Stability-Plasticity Dilemma in Outdoor Class-Incremental 3D Object Detection | 2026-11-02 | ShowClass-incremental learning (CIL) for outdoor 3D object detection remains largely underexplored. Existing outdoor studies are limited to single-benchmark evaluations, with SDCoT++ considering KITTI and CL2OD considering nuScenes under different incremental settings. This makes it difficult to assess whether observed continual-learning behavior generalizes across autonomous-driving benchmarks. To the best of our knowledge, we provide the first unified cross-benchmark analysis of outdoor class-incremental 3D object detection across multiple autonomous-driving benchmarks, covering both two-task and multi-stage settings. We investigate: (1) how severely existing 3D detectors forget previously learned classes, (2) how incremental training exposes the stability-plasticity dilemma between knowledge retention and new-class acquisition, and (3) which forms of supervision are most effective for balancing these objectives. We introduce CL3DOD, a class-incremental 3D detection framework that restores missing supervision for previous classes using predictions from the previous-stage detector, preserves old-class responses through query-wise knowledge distillation, and employs EMA-based temporal consistency regularization. Experiments on KITTI and nuScenes reveal a consistent stability-plasticity dilemma across both benchmarks. CL3DOD generally achieves a more favorable stability-plasticity balance than controlled incremental baselines and improves upon published prior methods in the reported benchmark comparisons. This advantage persists across multiple incremental stages, demonstrating strong retention of previously learned knowledge while maintaining effective new-class acquisition. Our code, evaluation setup, and incremental task splits will be publicly released. |
|
| Towards Continual Motion-Language Agents: LoRA Variants for Incremental Motion Understanding and Generation | 2026-11-02 | ShowMotion-language agents must possess the bidirectional capability to both understand human movement (motion-to-text, M2T) and generate it from natural language (text-to-motion, T2M). While foundational models have achieved strong performance in static settings, autonomous agents operating in dynamic environments must continuously incorporate new motion concepts -- such as novel athletic styles or specialized gestures -- without catastrophic forgetting of previously acquired skills. We investigate the stability-plasticity trade-off in bidirectional motion-language learning under sequential task exposure. Building on a frozen large language model backbone, we introduce low-rank adaptation (LoRA) variants designed to mitigate inter-task interference. We specifically propose mixture-of-experts architectures that utilize an autoencoder-based router to select task-specific experts at inference time, so that no task-label is needed. To evaluate these methods, we establish a reproducible five-task benchmark derived from HumanML3D through semantic clustering of motion descriptions. Our experimental results demonstrate near-zero forgetting across both M2T and T2M directions while maintaining high generation and captioning quality. Furthermore, we show that hard expert selection via routing significantly outperforms soft expert blending in quality metrics, indicating that preserving expert isolation is critical for maintaining performance in our continual learning setting. Finally, we observe that a divergence between token-level accuracy and downstream generation quality may occur, highlighting the need for more comprehensive evaluation protocols in future research on lifelong motion-language agents. |
19 pa...19 pages, 1 figure. Updated author version of the paper published at CoLLAs 2026. Adds merged-adapter analysis and comparative evaluations, clarifies routing and interpretation of results, and integrates the approach overview into the main text |
| CARE: Grounded Recursive Self-Improvement for Production Customer-Service Agents | 2026-11-02 | ShowRecursive self-improvement (RSI) in production agents involves transforming accumulated experience into reliable model updates while adapting to evolving business policies. We present CARE (Customer-service Agent Retrospective and Evolve), a grounded RSI framework that operationalizes this process in customer service. Its goal is to replace a frontier proprietary incumbent with an in-house mid-sized model that continues learning after deployment. After a one-time cold start on governed production traces, the current model serves production traffic alongside the incumbent, and its own failures drive the next update. A 24/7 inspection process establishes failures through evidence-blinded, independently reviewed judgments. Attribution separates model errors from defects in Skills, tools, or the harness, routing component failures to repair. Model-attributable trajectories are truncated at the first erroneous decision and paired with verified repairs. On-policy self-distillation (OPSD) then updates the model using the same model as teacher, with the diagnosis and repair supplied as privileged information unavailable to the student. To support continued evolution as business requirements change, training mixes failures from different Skill/SOP versions, each judged against the policy in force when it occurred. Candidates run an executable simulation test compiled from the latest SOP and then frozen, followed by offline inspection of the resulting traces. Promotion requires a combined simulation-and-inspection score at least at parity with the deployed version, confirmation through an online comparison with the incumbent, and human approval. |
|
| InTAct: Interval-based Task Activation Consolidation for Continual Learning | 2026-11-01 | ShowContinual learning seeks to acquire new knowledge while preserving previously learned representations. Despite the success of prompt-based methods, they remain fragile in domain-incremental learning scenarios, where shifts in the input distribution cause representation drift in shared layers and lead to forgetting. We introduce InTAct, a method that preserves functional behavior in shared representations without freezing parameters or retaining past data. InTAct constrains updates within activation regions associated with prior domains, while allowing flexible adaptation elsewhere, thereby stabilizing neuron functionality rather than directly restricting parameter values. The approach is architecture-agnostic, integrates seamlessly with existing prompt-based frameworks, and matches or outperforms state-of-the-art methods on domain-incremental benchmarks. |
Accep...Accepted to NeurIPS 2026 (main track) |
| HQNN-CL: Regime-Aware Hybrid Quantum Neural Network with Continual Learning for Financial Forecasting | 2026-11-01 | ShowFinancial time-series forecasting is still difficult owing to the non-stationary nature of the market, regime changes, and catastrophic forgetting, which makes conventional deep learning models hard to adapt in the long term. To achieve uncertainty-aware robust financial forecasting in the face of changing market conditions, this study proposes a Hybrid Quantum-Classical Neural Network with Continual Learning (HQNN-CL). The proposed framework combines Gaussian Hidden Markov Model (HMM) based regime detection, BiLSTM temporal encoding, adaptive-depth parameterized quantum circuits, Elastic Weight Consolidation (EWC), experience replay, and probabilistic forecasting in a single architecture. The framework was tested on four real-time stock datasets: Microsoft, Systems Limited, Teradata and Tata Consultancy Services, and compared with seven state-of-the-art forecasting models, including recurrent and Transformer-based models. The experimental results show that HQNN-CL has outperformed other methods on the Systems Limited (RMSE = 3.0149) and Teradata (RMSE = 0.8909) datasets and is within 0.32% and 0.40% of the best baseline method for Microsoft and Tata Consultancy Services, respectively. The proposed framework also generates well-calibrated probabilistic forecasts within the range of 89.5% to 96.5% empirical coverage rate, has zero catastrophic forgetting (BWT = 0.0000) for all sequential learning tasks, and consistently shows positive forward transfer. Ablation and statistical significance analyses also validate the performance of adaptive quantum feature learning and continual adaptation under heterogeneous market conditions, regime-wise. |
20 Pages, 8 Figures |
| Can Large Language Model Agents Learn by Watching Tutorial Videos? | 2026-11-01 | ShowHumans continuously acquire knowledge from carefully crafted materials such as books and tutorial videos, carrying that understanding into future tasks. Millions of such instructional videos are uploaded online every day, yet LLM agents remain unable to exploit this ever-expanding wealth of human teaching. Their capabilities are frozen after training, while parameter updates at scale remain prohibitively expensive. This raises the central question: can LLM agents learn from this vast instructional resource without parameter updates? We introduce Tutorial Agentic Learning (TAL), a nonparametric continual-learning paradigm that consolidates tutorial videos into a persistent, structured tutorialbook memory through a two-stage adversarial process that harvests hard, video-grounded questions and uses them to refine the memory. At inference time, TAL retrieves through a Table-of-Contents (TOC) tree, tying retrieval directly to the model's reasoning rather than to embedding similarity. To study this setting, we introduce VidLearn, a multimodal QA benchmark built from 72 chess tutorial videos spanning 12 openings. On VidLearn, TAL outperforms all evaluated baselines and improves monotonically as more videos are consumed, achieving up to 113% and 91% relative gains over the corresponding vanilla MLLMs on the training and test splits, respectively. In actual gameplay, TAL reaches a stronger opening-phase position than its vanilla counterpart in 69% of head-to-head games. Evaluation in a second domain further demonstrates the generality of TAL. Together, these results highlight instructional videos as a scalable resource for parameter-update-free agent continual learning. |
|
| Mitigating Catastrophic Forgetting in Task-Free Federated Continual Forecasting via Two-Tier Replay under Concept Drift | 2026-10-31 | ShowFederated time-series forecasting is increasingly important in applications where temporal data are distributed across privacy-sensitive clients, such as environmental monitoring, smart cities, and industrial sensing. A central limitation of current continual-learning protocols is that they usually assume predefined task boundaries, whereas deployed forecasting systems receive an uninterrupted stream and must decide autonomously when adaptation is necessary. This paper addresses catastrophic forgetting in task-free federated forecasting under concept drift. We introduce a drift-aware continual-learning protocol in which each client monitors prediction-window RMSE, calibrates a robust median threshold from the offline base period, and triggers adaptation only after persistent degradation. To mitigate forgetting, we propose a TwoTierReplay mechanism that separates fixed offline representative memory from online memory. The offline tier preserves long-term knowledge learned during the base phase, while the online tier captures recent distributional changes after drift. Experiments on the Beijing Multi-Site Air- Quality dataset with 12 clients, three forecasting targets, and three horizons show that TwoTierReplay achieves the best stability-plasticity score in all nine target-horizon configurations. It improves over the strongest baseline by approximately 11.4%. These results suggest that explicit temporal structuring of replay memory is a practical and effective mechanism for task-free federated continual forecasting. |
Peer-...Peer-reviewed and accepted for presentation at the AGCS 2026 conference |
| Skill++: Towards Lifelong Agents via Routing-Aware Skill Refinement | 2026-10-31 | ShowLarge language model (LLM) agents extend their capabilities with skills, each with a name, a description, and a body. Under progressive disclosure, an LLM router selects skills by name and description alone and loads only the selected bodies. Skill libraries grow continually as users and agents add new skills. Because the router relies on this metadata, a similar new skill can displace an earlier one, breaking previously correct routing even though the earlier skill and the router are unchanged. This routing interference is aggravated for public skills, whose descriptions may not match their bodies. We formulate the problem as lifelong skill routing, a class-incremental continual learning problem. A solution must keep earlier skills selectable without suppressing new ones. Existing approaches fall short, since continually retraining the router risks catastrophic forgetting and rewriting methods either ignore similar skills or require per-skill feedback. We therefore propose Skill++, which rewrites only the name and description of each incoming skill, without router training or labeled data. The new description states what the body supports, how the skill differs from its closest predecessor, and what related work it excludes. As a library grows to 731 skills over ten sessions, Skill++ selects all required skills for 76.3% of queries, versus 36.4% under the original metadata. Across three benchmarks and five LLM routers, it outperforms all compared approaches, and its gains carry over to the Codex CLI agent harness. |
|
| Exo: An Architecture for Self-Improving Systems at Runtime | 2026-10-30 | ShowAgent harnesses govern how language models act in an environment; designing their policies by hand is difficult, while automated techniques require representative datasets and costly executions. We present Exo, an agent runtime architecture for safe learning during ongoing work that collapses the optimization loop into task execution, allowing agents to revise both their harness and their strategies for further improvement. A minimal protected mechanism layer preserves execution history and governs policy activation while the harness evolves through inspectable code changes. On long-running objectives in an open-world game, Exo reduces model cost by 12-53% relative to Codex and Pi in standalone runs and 45-82% in transfer runs, while matching task completion rates. Compared with the same starting policy without the ability to edit policy, Exo reduces held-out model cost by 35-61% on AppWorld and SREGym with comparable correctness, cutting AppWorld's mean task time by 13% and SREGym's median diagnosis time by 35%. On Continual Learning Bench coding tasks, model cost including learning falls by 14%, with both policies resolving every issue. |
|
| Implicit Assumptions About Tasks in Continual Learning | 2026-10-30 | ShowContinual learning is the study of how a model accumulates knowledge from data received over time, rather than all at once. In practice, this is typically achieved by dividing a dataset into tasks and presenting them sequentially. Explanations of the resulting measurements take into account factors such as the similarity between segments, the rate at which the data changes and the order of presentation. Each of these factors can be defined in terms of the underlying data, the partition imposed on it, or the learner that encounters it. These levels do not necessarily coincide, and it is usually left implicit which level a reported quantity refers to. We separate these and identify which properties acquire a value at each level. In controlled teacher--student experiments, we demonstrate that partitioning the same underlying trajectory at different levels of granularity, changing the order of presentation, and varying the similarity structure of the underlying data all result in different measured outcomes. We review representative benchmarks to determine which of these properties are stated. Where the underlying data cannot be inspected, as in foundation models built on undisclosed pre-training corpora, its properties are inherited by the model's initial state, and the training applied on top of it can be described. |
Accep...Accepted to the Workshop on Continual Learning in the Era of Foundation Models and Embodied Agents at NeurIPS 2026 |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| FLEG: Feed-Forward Language Embedded Gaussian Splatting from Any Views via Compact Semantic Representation | 2026-11-01 | ShowWe present FLEG, a feed-forward network that reconstructs language-embedded 3D Gaussians from multi-view images. Previous feed-forward language-embedded Gaussian reconstruction methods are restricted to a fixed number of input views and typically attach a language-aligned semantic embedding to each Gaussian, resulting in impractical input settings and semantic redundancy. In contrast, we introduce a geometry-semantic dual-branch distillation framework that supports a variable number of uncalibrated and unposed multi-view images as input. We also propose a novel-view-based distillation strategy during training that mitigates overfitting to input views. In addition, we observe that semantic representations are significantly sparser than geometric ones, and per-Gaussian language embedding is unnecessary. To exploit this sparsity, we design a decoupled language embedding strategy that represents language information with a sparse set of semantic Gaussians, rather than attaching embeddings to every Gaussian. Compared with dense pixel-aligned per-Gaussian embedding schemes, our method uses only 5% of the language embeddings while maintaining comparable semantic fidelity, effectively reducing storage costs. Extensive experiments demonstrate that FLEG outperforms state-of-the-art feed-forward reconstruction and language-embedded Gaussian methods in both reconstruction quality and language-aligned semantic representation. |
Accep...Accepted by ECCV 2026. Project Page: https://fangzhou2000.github.io/projects/fleg |
| Coarse3D: Controllable and Faithful Sparse Structure Generation for Image-to-3D | 2026-10-30 | ShowCoarse-to-fine 3D-native generation relies heavily on the first-stage sparse structure, which determines the global orientation and overall geometry for subsequent refinement. However, existing coarse generators suffer from two key limitations: their orientation is difficult to control due to ambiguous semantic canonicalization, and their structures are often inconsistent with input images due to weak 2D-3D grounding. We propose Coarse3D, a sparse-structure generation model that produces orientation-controllable and structurally faithful voxel geometry from arbitrary unposed images. Coarse3D introduces three designs: a first-frame anchored canonical space that aligns generation with the first input view while preserving object-centric normalization; a dual-tower DiT that couples generation with auxiliary feed-forward reconstruction to learn spatially grounded features; and spatial-prior guided attention that regularizes cross-attention with pixel-to-voxel correspondence. Coarse3D can directly replace the coarse stage of existing pipelines. Experiments show that it substantially improves sparse structure quality and consistently enhances final 3D generation, achieving state-of-the-art performance. |
|
| L3G-World: Real-Time Infinite World Generation with Latent 3D Gaussian Memory | 2026-10-29 | ShowInteractive world models generate visual content conditioned on user actions, yet maintaining spatial consistency over long-horizon exploration in real time remains challenging. Existing approaches rely on either temporal context or explicit 3D memory, both of which scale poorly because memory grows with the length of the generated stream. Moreover, explicit 3D memories condition latent video generators by re-encoding RGB renderings, which precludes real-time generation. In this paper, we introduce a Latent 3D Gaussian Memory (L3G) that combines the efficiency of feed-forward reconstruction with a compact, rapidly renderable scene representation. Building on this memory, we propose L3G-World, in which a memory updater and a video diffusion transformer (DiT) alternate autoregressively to generate spatially consistent worlds. The memory updater spawns Gaussians in newly observed regions and refines existing ones, so that the memory grows with scene extent rather than stream length. Each Gaussian further stores a VAE latent feature, allowing the memory to be rendered directly into the DiT's latent space without VAE re-encoding. Experiments show that L3G-World achieves state-of-the-art generation quality and long-horizon spatial consistency, with up to |
Proje...Project page: https://cvlab-kaist.github.io/L3G-World/ |
| SparseTokenGS: Gaussian-based Sparse-View CBCT Reconstruction via Reusable Token Pool | 2026-10-28 | ShowSparse-view cone-beam computed tomography is widely used for fast, low-dose volumetric imaging and lesion localization in dentistry and image-guided interventions, but it introduces severe artifacts. Feed-forward Gaussian methods transfer the reconstruction cost to training and decode each coordinate from projection-aligned features and neighboring anchors. However, information from the whole scan is written only when the field is constructed and cannot be retrieved during coordinate decoding, leaving the volume inconsistent within the scan. This is a common limitation of such methods: once the field is fixed, decoding uses only local features and reconstruction quality drops substantially. We propose SparseTokenGS, a Gaussian-based method for sparse-view cone-beam computed tomography reconstruction via a reusable token pool. Anchor Token Fusion and Point-Token Attention query this shared pool at the anchors and at the query coordinates, respectively. The local Gaussian prior is retained, and the volume is reconstructed in one forward pass, without per-scan optimization or a projection-domain loss. On three public chest and dental datasets, SparseTokenGS achieves the highest reconstruction performance among the compared methods. On clinical dental scans, reconstruction from only six projections reaches 29.51 decibels in peak signal-to-noise ratio and a structural similarity of 0.842 against the full-view reference, and off-the-shelf tooth-and-pulp segmentation keeps these structures separable for root-canal localization. Experiments on volcanic rock and chicken show performance after same-object adaptation. A reusable token pool gives feed-forward reconstruction a mechanism that keeps scan-wide context available through decoding and can benefit future rapid clinical lesion localization and analysis. The code is available at: https://github.com/FrankDengAI/SparseTokenGS. |
10 pa...10 pages, 12 figures. Submitted to IEEE Transactions on Medical Imaging |
| EUReCA: End-to-End Unified CBCT Reconstruction Across Clinical Acquisitions | 2026-10-27 | ShowCone-beam computed tomography (CBCT) is widely used for image-guided radiotherapy, but reducing the number of projections to lower dose and scan time makes reconstruction increasingly ill-posed. We present EUReCA (End-to-end Unified REconstruction across Clinical Acquisitions), a geometry-aware feed-forward model for variable-view clinical linac CBCT. For a given acquisition trajectory, EUReCA accepts 8-100 projections with a single set of weights, without view-count-specific retraining; the same architecture and training recipe extend unchanged to other clinical trajectories. It combines learned projection features with an in-frame Feldkamp-Davis-Kress (FDK) reconstruction. The model encodes each projection with a shared multiscale backbone, back-projects features into multiresolution 3D grids using the acquisition geometry, aggregates variable numbers of views at fixed computational cost, and refines the volume with a global 3D transformer and multilevel decoder. On simulated data, EUReCA achieved 29.9-33.2 dB PSNR with 10-100 views and outperformed all applicable feed-forward baselines; it also exceeded iterative optimization-based methods while running orders of magnitude faster. On 15 measured patient CBCT scans evaluated against vendor reconstructions, EUReCA achieved the best performance at every tested view count. These results suggest that geometry-aware, measurement-anchored feed-forward reconstruction can provide a practical and flexible foundation for variable-view linac CBCT. |
|
| RoGe: Novel View Synthesis via End-to-End Implicit Reconstruction and Generation | 2026-10-26 | ShowNovel view synthesis from sparse inputs requires both geometric grounding from the observed views and generative priors of unobserved regions, motivating recent hybrid methods that combine reconstruction and generation. However, existing methods bridge the two with rendered images or explicit 3D representations such as point maps or 3D Gaussians. Generation is thus conditioned on a lossy and imperfect projection of the scene, inheriting its errors, and reconstruction receives no signal from generation to correct them. We present RoGe, an end-to-end unified reconstruction and generation framework that removes this explicit bridge. It targets roaming within a scene anchored by sparse views: given a few posed images and a camera trajectory, it synthesizes a temporally coherent video along that trajectory. From the sparse input views, RoGe builds an implicit scene representation with a feed-forward reconstruction model, and queries it with camera rays to obtain per-view geometric features. These features are injected into a video diffusion model as conditioning, without any explicit 3D intermediate. Both modules are trained jointly, so the generation objective directly shapes its own geometric conditioning. We conduct extensive experiments, where RoGe outperforms reconstruction-based, generation-based, and hybrid baselines in terms of image-level quality and video-level temporal and geometric consistency. Ablations confirm that ray-queried implicit features outperform both raw reconstruction tokens and rendered RGB as conditioning, and that joint training brings further gains. Our code will be released on https://jerry-locker.github.io/roge/. |
|
| VGGT-$Ω$ | 2026-10-25 | ShowRecent feed-forward reconstruction models, such as VGGT, have proven competitive with traditional optimization-based reconstructors while also providing geometry-aware features useful for other tasks. Here, we show that the quality of these models scales predictably with model and data size. We do so by introducing VGGT-$Ω$, which substantially improves reconstruction accuracy, efficiency, and capabilities for both static and dynamic scenes. To enable training this model at an unprecedented scale, we introduce architectural changes that improve training efficiency, a high-quality data annotation pipeline that supports dynamic scenes, and a self-supervised learning protocol. We simplify VGGT's architecture by using a single dense prediction head with multi-task supervision and removing the expensive high-resolution convolutional layers. We also use registers to aggregate scene information into a compact representation and introduce register attention, which restricts inter-frame information exchange to these registers, in part replacing global attention. In this way, during training, VGGT-$Ω$ uses only about 30% of the GPU memory of its predecessor, allowing us to train with 15x more supervised data than prior work and to leverage vast amounts of unlabeled video data. VGGT-$Ω$ achieves strong results for reconstruction of static and dynamic scenes across multiple benchmarks, for example, improving over the previous best camera estimation accuracy on Sintel by 77%. We also show that the learned registers can improve vision-language-action models and support alignment with language, suggesting that reconstruction can be a powerful and scalable proxy task for spatial understanding. Project Page: http://vggt-omega.github.io/ |
CVPR 2026 (Oral) |
| FreeView: Robust 3D Reconstruction from Uncurated Views | 2026-10-24 | ShowReconstructing a coherent 3D scene from uncurated image collections remains challenging due to the presence of irrelevant or non-overlapping views. Recent work has shown that feed-forward reconstruction models contain internal view-consistency signals that support distractor rejection. However, reference-conditioned filtering relies on a curated clean reference, which is not guaranteed in practical image collections. To address this limitation, we introduce FreeView, a training-free framework that jointly selects a filtering reference and rejects distractor views from an uncurated collection. FreeView leverages collection-level consensus to identify a representative view that serves as a reliable reference for filtering. It then adapts view rejection to each input collection by combining complementary visual and geometric evidence. This approach preserves useful views under substantial appearance and viewpoint variations while rejecting distractors. FreeView integrates with both VGGT and pi^3, removing the need for a manually curated reference or manual calibration of view rejection. Extensive experiments across multiple datasets demonstrate high clean-view retention and distractor rejection rates, leading to reliable 3D reconstruction from uncurated views. |
|
| Graph3R: Unifying View Graph Initialization with Feed-Forward Reconstruction | 2026-10-17 | ShowFeed-forward reconstruction recovers camera poses and dense geometry from image sets whose views are assumed to come from a single scene. In practice, uncurated image collections often contain views from disjoint scenes, and reconstructing these views jointly can corrupt camera poses and geometry. Existing pipelines typically resolve view associations in separate stages before final reconstruction, leaving view graph initialization and reconstruction inference only loosely coupled. We introduce Graph3R, which unifies view graph initialization with feed-forward reconstruction through a shared geometric representation. A lightweight verification head predicts pairwise covisibility from the multi-view features of |
|
| Holi-Spatial 360: Generative Reconstruction from Sparse Panoramas for Spatial Rendering and Intelligence | 2026-10-14 | ShowSpatial intelligence requires scalable 3D environments that are both photorealistic and semantically grounded, yet existing indoor datasets are dominated by room-scale captures and costly LiDAR acquisition. Online real-estate platforms offer abundant panoramas of diverse multi-room homes, but sparse viewpoints and texture-poor regions make reconstruction severely under-constrained. We introduce Holi-Spatial 360, a generative reconstruction and data curation framework for converting these panoramas into annotated multi-room 3D environments. Because reconstruction quality bounds subsequent generation, our Correspondence-Guided Geometric Optimization converts feed-forward reconstruction priors into reliable cross-view constraints and builds a coherent 3D Gaussian Splatting representation as 3D spatial memory. Adapted single-image and video priors then restore appearance and complete continuous trajectories while preserving observed content. The reconstructed scenes support arbitrary RGB-D rendering and an automated curation pipeline that combines end-to-end 3D grounding with VLM-based verification. Holi360 contains over 1K multi-room 3DGS scenes, more than 100K instance annotations, and 2M spatial question-answer pairs. We also introduce the Holi360 benchmark to evaluate the spatial understanding of VLMs in real-world, multi-room indoor environments. Experiments demonstrate superior reconstruction and curation quality under sparse panoramic inputs, while fine-tuning VLMs on Holi360 substantially improves spatial reasoning capabilities. |
|
| Dense Dynamic Scene Reconstruction and Camera Pose Estimation from Multi-View Videos | 2026-10-05 | ShowWe address the challenging problem of dense dynamic scene reconstruction and camera pose estimation from multiple freely moving cameras -- a setting that arises naturally when multiple observers capture a shared event. Prior approaches either handle only single-camera input or require rigidly mounted, pre-calibrated camera rigs, limiting their practical applicability. We propose a two-stage optimization framework that decouples the task into robust camera tracking and dense depth refinement. In the first stage, we extend single-camera visual SLAM to the multi-camera setting by constructing a spatiotemporal connection graph that exploits both intra-camera temporal continuity and inter-camera spatial overlap, enabling consistent scale and robust tracking. To ensure robustness under limited overlap, we introduce a wide-baseline initialization strategy using feed-forward reconstruction models. In the second stage, we refine depth and camera poses by optimizing dense inter- and intra-camera consistency using wide-baseline optical flow. Additionally, we introduce MultiCamRobolab, a new real-world dataset with ground-truth poses from a motion capture system. Finally, we demonstrate that our method significantly outperforms state-of-the-art feed-forward models on both synthetic and real-world benchmarks, while requiring less memory. |
fix a...fix author name errors |
| Less Is More: Genetic Frame Selection for Efficient Novel View Synthesis | 2026-09-28 | ShowFeed-forward novel view synthesis reconstructs a scene from many input images in a single forward pass, yet more views do not necessarily improve performance: redundant or poorly chosen frames increase computational cost and may degrade reconstruction quality. We address the problem of selecting, from an already captured sequence, a fixed-size subset of input views that is most informative for reconstructing specified target viewpoints. We propose a render-free view selector that scores candidate frames based on three complementary criteria: target-view coverage, measured against observed frames that stand in for the targets, redundancy with previously selected views, and image sharpness. A lightweight scoring network then selects the most informative frames without rendering, reconstruction, or per-scene optimization at inference time. To train the selector, we distill an expensive offline search procedure in which a genetic algorithm identifies high-quality subsets by directly optimizing reconstruction performance on training scenes. The selector learns to reproduce these choices from geometric and image-level features alone. Across six datasets and multiple input budgets, our method consistently outperforms both geometric and reconstruction-aware view-selection baselines while incurring significantly lower selection costs than reconstruction-based alternatives. Moreover, carefully selected subsets can outperform feed-forward reconstruction from the full input sequence. The learned selector generalizes across diverse reconstruction paradigms (feed-forward, 3D Gaussian Splatting, and NeRF), to object-targeted reconstruction and to a cross-capture setting in which the target views come from a separate acquisition pass. More broadly, our results indicate that explicitly reasoning about target relevance and inter-view redundancy is a fundamental factor in efficient scene reconstruction. |
|
| From Feed-Forward to Flow: Unifying Reconstruction and Generation Is Easier Than You Think | 2026-09-26 | ShowReconstruct where the images provide evidence, and generate where they do not: recent success of spatial world models such as Atlas (World Labs Team, 2026) highlights the value of unifying reconstruction and generation in one model. Yet the two have long lived in separate paradigms with distinctive failure modes: feed-forward reconstruction averages ambiguity into blur, while conditional generation invents plausible but scene-inconsistent detail. In this work, we present a unified flow-based formulation for reconstruction and generation, where a shared clean-target predictor performs direct reconstruction at its single-step endpoint and unfolds conditional generation through multi-step flow. A controlled toy study reveals the mechanism: with a single step, the predictor collapses to the conditional mean just like feed-forward methods, favoring consistency over diversity. With multi-step inference, the fidelity of generated details grows with context richness: closer observations reduce ambiguity and yield better-matched details. We further instantiate the formulation in appearance and geometry 3D tasks. JiT-LVSM improves perceptual and distributional quality in novel view synthesis, while JUSt3R retains competitive single-step geometry prediction with additional multi-step inference capabilities that reduces veil and flying-pixel artifacts, producing cleaner surface structure with greater test-time compute. Together, they show that reconstruction and generation can share both a formulation and a backbone, with their behavior governed by denoising configuration---making unification surprisingly simple. |
34 pa...34 pages, including supplementary material. Haoru Wang and Qianfan Shen contributed equally |
| When Wider Views Fail: Stress-Testing Feed-Forward 3D Reconstruction | 2026-09-21 | ShowFeed-forward 3D reconstruction models enable efficient geometry estimation from sparse images, but their pretrained nature can make them vulnerable to distribution shifts beyond their training data. Identifying these failure modes is important for understanding when such models can be reliably deployed in unconstrained imaging settings. We investigate viewpoint variation as a controlled distribution shift by varying the angular span of sparse image inputs while keeping the input budget fixed. Across multiple feed-forward reconstruction models, we observe substantial degradation as viewpoint span increases, with wide spans producing both incomplete surface coverage and geometry unsupported by the observed imagery. These results reveal that viewpoint variation can induce failure modes beyond conventional reconstruction incompleteness, highlighting the need to evaluate pretrained feed-forward models under distribution shifts that challenge their learned geometric priors. |
|
| SURE-Map: Self-Correcting Streaming Geometric Foundation Models | 2026-09-21 | ShowStreaming geometric foundation models are emerging as a compelling alternative to SLAM systems. Yet this streaming nature introduces a fundamental issue: each prediction is made from limited context, which is vulnerable to dynamic objects and weak textures. Small local errors accumulate into severe geometric distortion and long-horizon scale drift. We argue that reliable streaming reconstruction requires geometric foundation models to be not only predictive, but also self-correcting. We introduce SURE-Map, a self-correcting framework built upon two complementary principles. First, we explicitly model cross-view geometric uncertainty. Unlike conventional depth or point confidence, which primarily reflects the reliability of individual-view prediction, our uncertainty directly measures whether the jointly predicted pose and depth induce geometrically consistent cross-view pixel correspondences. Second, because local correction alone cannot eliminate slowly accumulating scale errors, we introduce multi-timescale self-correction: fast consecutive-frame inference preserves streaming efficiency, while sparse keyframe-window inference provides longer-range geometric evidence to periodically recalibrate the scale of recent trajectories. SURE-Map establishes new state-of-the-art performance for online feed-forward reconstruction across long-horizon benchmarks, reducing ATE-RMSE from 24.00 to 17.24 m on KITTI, 5.11 to 4.74 m on Oxford Spires, and 31.37 to 28.58 m on VBR, with further improvements to 15.17, 4.63, and 22.12 m when incorporating loop-closure refinement. Project page: https://mingkai-liu.github.io/projects/sure-map/. |
Corre...Corrected a typo in the title; manuscript content unchanged |
| RegVGGT: Sustainable Visual Geometry Grounding for Streaming via Regulated Memory | 2026-09-20 | Show3D reconstruction from a lengthy video stream input poses a dilemma for feed-forward reconstruction models (FFRMs), that a whole-stream inference context cannot be retained under limited GPU memory.Recent studies seek to resolve this problem via a trade-off between the integrity of inference context and GPU memory usage, which either suffer from a rapid memory inflation or degraded context integrity due to artificially capping memory usage.Driven by our key observation that the initial saliency of a token reliably dictates its long-term importance across the stream, we propose RegVGGT, a training-free token regulation method which aggressively regulates the tokens of incoming frames.By admitting at most 1% of tokens per frame to update the context memory, our method dramatically suppresses memory inflation as the stream progresses.Equipped with a FlashAttention-compatible token saliency estimation scheme, RegVGGT is capable of processing thousands of frames on a consumer-grade GPU with negligible compromise to reconstruction quality.Extensive experiments demonstrate that RegVGGT achieves state-of-the-art performance on long-horizon benchmarks across diverse FFRM prediction tasks, surpassing prior FFRM-based stream reconstruction baselines by a large margin. |
ECCV ...ECCV 2026. Corresponding author is Junjun Jiang |
| GRF-Recon: Global Ray-Field Optimization for Long-Sequence Feed-forward Reconstruction | 2026-09-17 | ShowFeed-forward 3D reconstruction provides an efficient paradigm for scene modeling from image sequences. Scaling these models to large monocular scenarios are constrained by excessive GPU memory footprint, degraded local geometry, and long-term trajectory drift. Existing chunk-based optimization strategies provide limited geometric constraints and fail to maintain global consistency over extended trajectories. We present a unified framework for stable and scalable feed-forward 3D reconstruction from long monocular sequences. Our approach builds on coarse-to-fine trajectory alignment augmented by lightweight geometric prior injection. Distilling monocular geometric cues into the feed-forward backbone via LoRA adaptation improves depth accuracy on fine structures while preserving inference efficiency. We introduce a hybrid-weight sparse ray-field optimization that leverages high-frequency geometric features to guide local point-cloud refinement and enforce consistent inter-frame ray constraints. Unlike prior chunk-based methods, this establishes strong cross-frame geometric coupling while maintaining scalability. Finally, an efficient trajectory stitching strategy with joint ray-error optimization explicitly reduces accumulated drift. Extensive experiments show that our approach achieves competitive trajectory accuracy compared with representative SLAM systems, while maintaining globally consistent 3D reconstruction in large-scale scenarios. |
Accep...Accepted to ECCV 2026 as a Spotlight presentation |
| RoMa-$Ω$: What Feed-Forward 3D Models Know About Image Matching | 2026-09-08 | ShowLearned image matching has experienced significant progress in recent years, culminating in robust and accurate matchers such as RoMa, whose robustness is often attributed to its use of frozen DINO features. In a parallel development, feed-forward reconstruction models, such as VGGT, have been trained on ever-growing datasets to accurately regress dense 3D point maps and camera poses. The distinction between matchers and feed-forward reconstruction models has become increasingly blurred with the introduction of matching losses in models such as MASt3R and VGGT-$Ω$. This raises a natural question: what do feed-forward 3D models know about image matching? In this work, we answer this question by analyzing three scenarios: (i) zero-shot matching of patch features, (ii) direct matching of 3D point predictions, and (iii) training a full matcher on top of the learned representations. We find that, despite performing poorly in zero-shot matching, especially in later layers, feed-forward reconstruction models provide strong representations for linear probing and full matching pipelines. We further show that, even without any training, their raw predictions alone enable competitive matching, albeit only under moderate viewpoint changes and modality gaps. Based on these insights, we retrain RoMa v2 by replacing its DINO backbone with VGGT-$Ω$. Our resulting model, \ours, outperforms state-of-the-art matchers on a wide range of benchmarks, e.g. +8.1 mAA compared to RoMa v2 on WxBS. |
|
| AdaptSplat: Adapting Vision Foundation Models for Feed-Forward 3D Gaussian Splatting | 2026-09-06 | ShowThis work explores a simple yet powerful lightweight adapter design for feed-forward 3D Gaussian Splatting (3DGS). Existing methods typically apply complex, architecture-specific designs on top of the generic pipeline of image feature extraction |
|
| Sparse auto-regressive modeling for scene generation from multi-view images | 2026-09-03 | ShowGenerating complete 3D scenes from sparse, unconstrained views is a fundamental challenge in 3D vision which requires reasoning beyond observed content while remaining computationally tractable. Existing feed-forward reconstruction methods are inherently limited to content visible in the input images, while 3D generative modeling is hindered by the high computational cost of dense volumetric representations and the scarcity of large-scale 3D supervision. We introduce SPAR3S, a sparse voxel-aligned 3D latent generative model for conditional scene completion without requiring ground-truth 3D data for supervision. Our key insight is to formulate 3D scene generation in a structured, compact, voxel-aligned 3D latent space where only occupied voxels are represented. We learn this sparse latent space directly from multi-view images using photometric supervision via differentiable 3D Gaussian Splatting. Given a partial set of observed voxels encoded from sparse input views, scene completion reduces to predicting the missing latent tokens and their spatial support within the voxel grid. To this end, we train a masked autoregressive transformer that jointly models voxel occupancy and latent token values, enabling efficient and spatially consistent generation of unseen regions. We demonstrate the effectiveness of our method on synthetic indoor scenes, achieving higher novel-view quality than prior work. We further validate its generalization on RealEstate10k, highlighting its applicability to real-world data. |
Accep...Accepted at ECCVV 2026 |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| What Should a Navigator Remember? Spatiotemporal Memory for Zero-Shot Vision-Language Navigation | 2026-11-03 | ShowZero-shot vision-and-language navigation (VLN) requires an agent to use past observations without navigation-specific training. We study what information an 8B-scale navigation system should retain and how it should be organized for subsequent decisions. Our agent combines instruction progress, hierarchically compressed observation history anchored to poses, and accumulated geometry in complementary task, temporal, and spatial memories. An 8B observer summarizes observations, and an 8B reasoner uses the memories to select waypoint intents or request additional views, a geometric controller executes the selected targets. A semantic information perspective motivates the distinction between retaining evidence, making it accessible to the reasoner, and acquiring further observations. On R2R-CE val-unseen, the complete system achieves the highest among the evaluated Qwen3-based zero-shot baselines. Removing temporal memory reduces SR and sharply increases average path length from 16.9 to 39.1m, removing the spatial-map and planning component reduces SR as well. Real-world experiments on a Unitree Go2 show that our zero-shot method achieves higher success rates than the post-trained baseline in both corridor and office environments, suggesting better transfer to the evaluated real-world settings. As a complementary evaluation, LAW2 uses paired instruction interventions to examine behavioral responses, instruction following ability beyond endpoint success. These results support preserving task progress, observation history, and geometry in structured memory for zero-shot navigation with smaller open models. |
20 pa...20 pages, 3 figures, 8 Tables |
| HaltNav: Reactive Visual Halting over Lightweight Topological Priors for Robust Vision-Language Navigation | 2026-11-02 | ShowVision-and-Language Navigation (VLN) is shifting from rigid, step-by-step instruction following toward open-vocabulary, goal-oriented autonomy. Achieving this transition without exhaustive routing prompts requires agents to leverage structural priors. While prior work often assumes computationally heavy 2D/3D metric maps, we instead exploit a lightweight, text-based osmAG (OpenStreetMap Area Graph), a floorplan-level topological representation that is easy to obtain and maintain. However, global planning over a prior map alone is brittle in real-world deployments, where local connectivity can change (e.g., closed doors or crowded passages), leading to execution-time failures. To address this gap, we propose a hierarchical navigation framework HaltNav that couples the robust global planning of osmAG with the local exploration and instruction-grounding capability of VLN. Our approach features an MLLM-based brain module, which is capable of high-level task grounding and obstruction awareness. Conditioned on osmAG, the brain converts the global route into a sequence of localized execution snippets, providing the VLN executor with prior-grounded, goal-centric sub-instructions. Meanwhile, it detects local anomalies via a mechanism we term Reactive Visual Halting (RVH), which interrupts the local control loop, updates osmAG by invalidating the corresponding topology, and triggers replanning to orchestrate a viable detour. To train this halting capability efficiently, we introduce a data synthesis pipeline that leverages generative models to inject realistic obstacles into otherwise navigable scenes, substantially enriching hard negative samples. Extensive experiments demonstrate that our hierarchical framework outperforms several baseline methods without tedious language instructions, and significantly improves robustness for long-horizon vision-language navigation under environmental changes. |
|
| Beyond Trial-and-Error: A Geospatial-Guided Active Reasoning Framework for Robust Aerial Vision-and-Language Navigation | 2026-10-30 | ShowAerial Vision-and-Language Navigation (VLN) requires UAVs to locate language-specified goals in large-scale urban environments, where objects often appear small and visually ambiguous from high-altitude viewpoints. Existing methods largely follow a reactive paradigm common in indoor VLN, repeatedly updating navigation states from local observations while underutilizing the geospatial structure that remains stable during navigation. As a result, visually similar distractors can cause a reactive goal search to repeatedly reinforce locally plausible goal hypotheses based on passively accumulated evidence, leading to unreliable localization. To address this, we propose GeoReasoner, a robust framework that shifts the paradigm of Aerial VLN from reactive trial-and-error to geospatial-guided active reasoning. Specifically, GeoReasoner first performs pre-search geospatial reasoning to derive region-level priors and constrain the goal hypothesis space. Building on these, it further conducts hypothesis-driven exploration to actively acquire discriminative evidence and refine competing goal hypotheses. Finally, consistency-aware goal verification adversarially assesses complementary evidence sources before stopping, hence reducing premature or erroneous goal confirmations. Experiments on the real-world CityNav benchmark demonstrate that GeoReasoner sets a new SOTA among zero-shot Aerial VLN methods while competing favorably with supervised approaches. |
|
| MOON: Learning When to Forget and How to Move On in Vision-and-Language Navigation | 2026-10-30 | ShowRecent vision-and-language navigation (VLN) policies act on their visual history, and an incorrect action can take them off the instructed route. Paired interventions on four VLN policies show that clearing the history doubles pooled success at the onset of a deviation but lowers it on route, while later in a deviation, removing only the off-route frames helps much less. After a deviation, the retained history keeps the policy from turning where a turn is needed, which we call behavioral inertia. We propose MOON, a framework that equips a frozen VLN policy with two lightweight modules, adding less than 1% of its parameters. A deviation judge decides when to forget, raising an alarm that clears the history when it predicts that the agent is off route. A recovery policy, trained with GRPO from alarm states, learns how to move on, turning more of the successes found only by sampling into a single greedy rollout. Experiments on R2R-CE and RxR-CE show that MOON reaches the highest R2R-CE success rate among the compared methods and raises the success rate of its frozen base by 5.2 and 6.0 points, and an ablation with a random trigger indicates that this gain depends on when the history is cleared. Without retraining, MOON also improves zero-shot object-goal navigation on HM3D-OVON. Together, these results suggest that deviation recovery depends on knowing when to forget, not only how to act. Code is available at https://github.com/AutoLab-SAI-SJTU/MOON. |
|
| MATCH: Execution-Aware Hierarchical Reinforcement Learning for Vision-and-Language Navigation in Continuous Environment | 2026-10-28 | ShowVision-and-Language Navigation in Continuous Environments (VLN-CE) requires instruction-consistent planning and reliable execution under collision constraints. Yet many hierarchical approaches optimize the planner while keeping execution fixed. We introduce MATCH, a framework that jointly learns a language-conditioned planner and a goal-conditioned primitive-action executor. An execution-aware topological graph represents start-dependent execution options and integrates executor value estimates with realized outcome memory, enabling the planner to select execution requests and refine their targets through bounded continuous offsets. A transition-aware dense macro reward balances ordered progress along the reference route against execution residual, encouraging executable requests without substituting local success for instruction following. A staged curriculum culminates in synchronized on-policy updates, with the executor learning on planner-generated requests and the planner adapting its requests to the executor's current behavior. With 0.5B parameters, MATCH achieves state-of-the-art performance on the R2R-CE and RxR-CE benchmarks. Preliminary real-robot deployments further demonstrate its practical potential. |
|
| Goal2Pixel: Grounding Goals to Pixels for Vision-Language Navigation | 2026-10-27 | ShowVision-language models (VLMs) have become a common foundation for vision-and-language navigation in continuous environments (VLN-CE). Yet most VLM-based methods cast navigation as low-level action prediction, an interface that is ambiguous, tied to short-horizon motion primitives, and inefficient due to repeated VLM querying. We propose Goal2Pixel, a pure pixel-based paradigm that reformulates VLN-CE as navigable pixel grounding. Rather than predicting actions, Goal2Pixel uses the image plane as a unified spatial interface between VLM reasoning and robot motion: the model predicts a visible navigable pixel, which is back-projected into a 3D waypoint for forward navigation. Non-forward actions are encoded by auxiliary left/right/bottom regions for turning and stopping. To support long-horizon navigation, we propose ViKeyMem, a visibility-aware keyframe memory, achieving a favorable performance-efficiency trade-off while reducing training cost by over 50%. To adapt pretrained VLMs to navigable pixel grounding, we introduce semantic embeddings and coordinate-aware auxiliary losses. Among methods trained without external data, Goal2Pixel achieves competitive state-of-the-art performance on R2R-CE and RxR-CE. On R2R-CE Val-Unseen, it achieves 54.1% SR and 52.5% SPL with only 7.75 VLM calls per episode. Compared with direct action prediction, Goal2Pixel improves SR by 21.2 percentage points while requiring 6 times fewer VLM calls. Project Page: https://baobao0926.github.io/Goal2Pixel/. |
8 pages |
| Asking Less, Adapting Better: Feedback-Efficient Test-Time Adaptation for Vision-and-Language Navigation | 2026-10-22 | ShowVision-and-Language Navigation (VLN) requires agents to follow instructions and navigate to target locations. However, current VLN agents often suffer severe performance degradation in unseen environments due to distribution shifts. Test-Time Adaptation (TTA) offers a promising paradigm to bridge this gap, yet existing methods that rely on unsupervised entropy minimization can propagate errors through the trajectory, causing cascading failures in navigation. Recent methods mitigate this by incorporating human feedback, but they apply trajectory-level supervision uniformly across all steps without localizing the actions is truly informative, leading to inefficient feedback utilization and mismatched optimization granularity. To address these limitations, we propose FETA, a Feedback-Efficient Test-Time Adaptation method that improves both feedback efficiency and optimization reliability. To reduce unnecessary human interventions, we introduce a hierarchical sample selection strategy to localize informative samples. Specifically, we first filter trajectories whose normalized policy entropy exhibits large fluctuations. Within each selected trajectory, we further select the critical action via an entropy-gradient saliency criterion and query a single binary feedback on its correctness. A discrepancy-aware entropy optimization objective then updates the policy only at the queried step, with adaptive scaling that assigns larger relative weights to samples with greater prediction-feedback discrepancy. Extensive experiments across multiple benchmarks demonstrate that FETA achieves state-of-the-art performance with fewer human interventions, while remaining robust under noisy feedback and continual adaptation settings. |
|
| AgriVLN: Vision-and-Language Navigation for Agricultural Robots | 2026-10-21 | ShowAgricultural robots are serving as powerful members across a wide range of tasks, nevertheless, they are still heavily relying on manual operations or railway systems for movement. In contrast, Vision-and-Language Navigation (VLN) enables navigating robots from starting points to target positions following natural language instructions, having shown strong performances across various domains. However, none of the existing benchmarks or methods are specifically designed for agricultural robots. To bridge this gap, we build our A2A benchmark, which consists of 1,560 episodes across 6 classes of agricultural scenes using the step-by-step format for instructions and the continuous environment for actions. We evaluate several existing state-of-the-art VLN methods on A2A, then we observe that they perform well on short instructions while perform poorly on long instructions. To alleviate this phenomenon, we design our STL module, which decomposes a complex instruction into a list of simple subtasks, thereby assisting the decision-maker to focus on the current short subtask rather than the entire long instruction at any time. We integrate STL into the NavGPT backbone to propose our AgriVLN method, achieving a sufficient improvement on SR from 0.33 to 0.47 with a slight sacrifice on NE from 2.76 m to 2.91 m. We further implement the ablation studies and the case study, discussing the effectiveness and limitations of AgriVLN on different subtask quantities and scene classes. Repository: https://github.com/AlexTraveling/AgriVLN. |
|
| ReInsVLN: Revisiting Historical Evidence for Training-Free Vision-and-Language Navigation | 2026-10-19 | ShowVision-and-language navigation (VLN) relies on retaining and using relevant observations. Yet growing histories introduce redundancy, and retained information may remain underused. We introduce ReInsBench, a diagnostic benchmark that examines history utilization through controlled continuation trials. Its results show that history-review reinstruction can recover navigation failures using the existing task context. Motivated by this finding, we propose ReInsVLN, a training-free framework supporting memory construction and utilization in frozen navigators. Coverage- and Instruction-Aware Compression constructs compact visual context through temporal redundancy reduction, coverage-aware protection, and instruction-guided token selection. Forgetting-Inspired History Reminders guide navigators to revisit retained observations according to a predefined schedule, without additional model calls. On R2R Val-Unseen, ReInsVLN improves success rates by 8.48 and 3.37 percentage points over the native StreamVLN and InternVLA-N1, reaching 64.27% and 64.44%, respectively. Visual compression largely preserves navigation performance while reducing prefill latency and KV cache memory by up to 21.77% and 32.25%, respectively, on matched observation workloads. Real-world trials on humanoid and quadruped robots further support the applicability of ReInsVLN to sequential instruction following and ordinal navigation. |
8 pag...8 pages, 6 figures, 6 tables |
| Revealing and Bridging the Physical Executability Gap in Vision-Language Navigation | 2026-10-19 | ShowContinuous vision-and-language navigation (VLN) is typically evaluated by goal-reaching success, assuming it reflects an agent's ability to execute language-grounded behavior. We show that this assumption can fail: permissive contact models in current simulators may mask physically infeasible actions, creating a physical executability gap between semantic competence and embodied realization. To bridge this gap, we introduce Visual-Semantic Traversability Alignment (VeSTA), a framework that aligns VLN policies with physically executable futures. We define visual-semantic traversability as the ability to predict instruction-conditioned trajectories that are semantically valid, geometrically feasible, and physically robust. VeSTA addresses this misalignment through two complementary mechanisms: improving executable support in trajectory generation and risk-aware selection among multimodal futures. Specifically, Trajectory Relative Group Policy Optimization uses strict-contact feedback to reshape the candidate distribution toward executable regions, while the Traversability Risk Critic learns a risk-adjusted scalar utility for candidate-wise future selection. Experiments on R2R and RxR reveal substantial dependence of existing VLN agents on simulator contact semantics. Under strict-contact evaluation, VeSTA substantially reduces collision-induced failures while preserving or improving goal-reaching performance, and controlled real-world robot experiments provide evidence that the learned preference transfers beyond simulation. |
|
| SpaAct: Spatially-Activated Transition Learning with Curriculum Adaptation for Vision-Language Navigation | 2026-10-17 | ShowVision-and-Language Navigation (VLN) aims to enable an embodied agent to follow natural-language instructions and navigate to a target location in unseen 3D environments. We argue that adapting VLMs to VLN requires endowing them with two complementary capabilities for acquiring such awareness, namely backward action reasoning (why) and forward transition prediction~(how). Based on this insight, we propose SpaAct, a simple yet effective training framework that activates the dynamic spatial awareness in VLMs. Specifically, SpaAct introduces two spatial activation tasks: Action Retrospection, which asks the model to infer the executed action sequence from visual transitions, and Future Frame Selection, which forces the model to predict the visual transitions conditioned on history and action. These two objectives provide lightweight supervision on both backward action reasoning and forward transition prediction, encouraging the model to build dynamic spatial awareness in a VLM-friendly way. To further stabilize adaptation, we design TriPA, a Tri-factor Progressive Adaptive curriculum learning method that organizes training samples from easy to hard, allowing the model to gradually acquire navigation skills from basic locomotion to long-horizon reasoning. Experiments on standard VLN-CE benchmarks show that SpaAct consistently improves VLM-based navigation and achieves state-of-the-art performance. We will release the code and models to support future research. |
Accep...Accepted by ACM-MM 2026 |
| Embodied Agents Take Control: Minimal-Interface Zero-Shot Agents Rival Industrial-Scale Policies in Vision-and-Language Navigation | 2026-10-17 | ShowAutonomous embodied agents must sustain a long decision-making loop of perceiving, acting, verifying, and self-correcting. In trained end-to-end policies and hand-designed workflows, the interaction structure remains externally specified. We study how much of this loop a general-purpose model can direct through a minimal interface. We term this organization agentic embodied control: the reasoning model directly steers every action, keeping reasoning and control aligned. We evaluate fourteen frontier multimodal models as provided by their vendors, with monocular RGB and primitive actions across six navigation and embodied question-answering benchmarks. On 100-episode subsets, the best agents exceed published scores and approach or surpass our human tester. Who directs the loop matters: SmartWay and O2C-Nav using the same strong models achieve substantially lower success rates on R2R-CE. The lightweight mini-swe-agent harness performs comparably to vendor harnesses. Stronger models benefit from optional rather than imposed waypoints; adding SLAM raises R2R-CE success to 88%. Physical trials expose limitations in body awareness, execution feedback, and persistent spatial understanding. |
|
| SimpleVLN: Towards Simple and Frontier Vision-Language Navigation | 2026-10-16 | ShowVision-and-Language Navigation (VLN) requires embodied agents to ground language in visual observations and make long-horizon decisions. Existing methods often address these demands through task-specific architectures. As multimodal large language models (MLLMs) become increasingly capable, an important question remains: Have we truly unlocked the potential of native MLLMs for VLN before introducing task-specific architecture? We present SimpleVLN, an architecture-preserving framework developed through studies of navigation formulation, data construction, and optimization. At the formulation level, SimpleVLN combines single-turn decision inputs with fixed-budget global-local memory to efficiently use navigation history without unbounded context growth, while action chunks enable efficient output of multiple primitive actions per model invocation. For data construction, we combine expert demonstrations with policy-dependent DAgger experience and develop a scalable synthesis pipeline that generates large-scale tasks for reinforcement learning. At the optimization level, a progressive IL-DAgger-RL pipeline advances from action imitation to task-level outcome optimization. Comprehensive experiments validate the effectiveness of these three aspects. We instantiate SimpleVLN across multiple MLLM backbones and model sizes, demonstrating its generality. SimpleVLN achieves 76.7% and 80.1% success rates on the R2R-CE and RxR-CE splits, respectively, substantially outperforming existing methods. These results show that the navigation capabilities of native MLLMs can be effectively unlocked without task-specific architectural designs. |
homep...homepage: https://simplevln.github.io/ |
| PASS-VLN: Passability-Aware Local-Command Vision-and-Language Navigation | 2026-10-16 | ShowReliable vision-and-language navigation (VLN) requires not only understanding instructions but also determining whether and how a robot can physically execute them. However, existing VLN benchmarks typically assume executable instructions and focus on long-horizon tasks. This leaves body-dependent local execution, including passability and collision-free execution, largely unexamined. Evaluating these factors within long-horizon tasks would entangle local physical execution with task planning and long-term spatial memory, making navigation failures difficult to attribute. We therefore focus on grounding local navigation commands in executable motion under environmental and robot body constraints. We present PASS-VLN, a passability-aware local-command VLN benchmark and model. The benchmark holds the scene, instruction, start pose, and reference route fixed while varying robot body width, requiring target completion when the instructed passage is feasible and stopping before blockage otherwise. Our model encodes body width through obstacle inflation in a depth-derived local bird's-eye view, fuses geometry with visual and linguistic context, and predicts discrete latent action codes for continuous closed-loop execution. Experiments show that PASS-VLN improves collision-free success by approximately 12 percentage points over a strong VLN model fine-tuned on our benchmark, while adapting reach-versus-stop behavior to body constraints and generalizing to unseen body widths. When integrated with a high-level planner, PASS-VLN achieves higher success with fewer local commands, while real-world deployment further demonstrates its effectiveness as a low-level executor. The benchmark and model will be released. |
9 pages, 4 figures |
| BGC-Nav: Cross-Altitude Multi-UAV Collaborative Navigation via Belief-Guided Coupling | 2026-10-10 | ShowDespite recent advances in Aerial Vision-and-Language Navigation (VLN), single Unmanned Aerial Vehicle (UAV) search-and-rescue in outdoor scenarios remains challenging due to partial observability, making cross-altitude collaboration a promising direction. However, existing collaborative navigation methods rely on precise target localization and rigid goal-dispatching paradigms, ignoring high-altitude observation uncertainty and suppressing low-altitude exploration autonomy. To address these issues, we introduce BGC-Nav, a training-free cross-altitude collaboration framework that enables robust navigation via Belief-Guided Coupling (BGC). Specifically, the high-altitude UAV maintains a Global Belief Memory encoding task-conditioned priors, which the low-altitude agent incorporates into decision-making via Local Exhaustion Memory (LEM), Local Attraction Augmentation (LAA), and Directional Belief Prior (DBP) as an advisory bias rather than a rigid command. Extensive experiments on ESARBench show that BGC-Nav consistently outperforms state-of-the-art single-agent and collaborative methods. Code is available at https://github.com/uestchjw/BGC-Nav. |
|
| Successful but Unsafe: An Analytical Benchmark for Revealing Collisions in Vision-and-Language Navigation | 2026-10-10 | ShowVision-and-language navigation (VLN) agents are typically evaluated by whether they reach a target and how efficiently they follow an instructed route, yet these metrics provide little information about collision behavior. Moreover, continuous VLN evaluation commonly uses a sliding regime, where an agent can slide along obstacles after a collision, potentially masking the effect of unsafe actions on navigation performance. We introduce VLN-Safe, a benchmark for analyzing collision safety under two complementary evaluation regimes: sliding, reflecting common VLN evaluation practice, and blocking, where collision-inducing actions prevent forward motion and more directly expose their consequences. Across six published agents on R2R-CE, 38.5-56.8% of successful episodes contain contact under sliding, while switching to blocking decreases success rate by 11.7-36.1 percentage points. We further analyze signals for anticipating imminent collisions and find that depth geometry and recent movement history are more informative than policy-derived signals, while safe alternatives near the policy's preferred action are rarely available. Motivated by these findings, we introduce a training-free collision guard that vetoes risky forward actions, executes an escape maneuver, and returns control to the original policy. Under blocking, the guard reduces collision steps by 62-98% and improves success rate by 6.9-29.3 percentage points across five recurrent, topological, and vision-language-action agents. Under sliding, it leaves success rate unchanged while raising the share of collision-free successes by 4.1-12.4 points. These results reveal a substantial gap between task success and collision-safe navigation and show that lightweight collision-aware intervention can improve both safety and navigation performance. |
|
| GuidedNav: Shaping Vision-Language Navigation Representations through Pre-action Attention and Spatial Guidance | 2026-10-09 | ShowVision-and-language navigation (VLN) requires an agent to track progress through an extended instruction, identify landmarks relevant to its current objective, and understand the surrounding spatial structure to act. End-to-end navigation policies typically rely on action supervision to implicitly learn these capabilities. Without explicit guidance on the evidence underlying each decision, however, policies can overfit to correlations predictive of training actions rather than learn the task-relevant grounding and geometry needed for generalization. Our insight is to guide the representations that connect instructions to actions: identifying what remains to be done determines which visual evidence matters, while understanding the geometry around that evidence informs how to move. We introduce GuidedNav, a unified framework that explicitly shapes these representations through training-time guidance. Subtask and landmark guidance steer native pre-action attention toward the active instruction segment and relevant visual regions, grounding action prediction in the current objective. Spatial guidance aligns the policy's internal representations with geometric features that capture long-horizon context and local spatial structure. Together, these objectives strengthen navigation capabilities while avoiding the inference overhead of intermediate textual reasoning and online geometric encoding. We also introduce Guided-R2R and Guided-RxR, two annotation-enriched navigation datasets that support more interpretable VLN research. GuidedNav establishes state-of-the-art performance among the compared monocular RGB policies, improving success rates by 8.1% and 13.1% relative to the strongest prior results on R2R-CE and RxR-CE val-unseen, respectively, while achieving faster inference. Extensive experiments further demonstrate the effectiveness and real-world transferability of our approach. |
|
| Systematic Multi-Agent Vision-and-Language Navigation: Formulation, Benchmark, and Method | 2026-10-07 | ShowVision-and-Language Navigation (VLN) has largely focused on a single agent following a single instruction, yet many real-world applications require teams of robots to tackle tasks beyond the capabilities of any individual agent. We present Systematic Multi-Agent Vision-and-Language Navigation, providing, to our knowledge, the first systematic formalization of multi-agent VLN as a constrained coordination problem: each mission consists of subtasks carrying dependency and resource constraints (presence locks and holding chains). A verified four-stage crafting pipeline instantiates the task as MAVLN, comprising 11,724 episodes across 145 scenes with teams of up to four agents under three instruction regimes, accompanied by tailored constraint-aware metrics. We further present TRISS, a coordination-ready navigation system coupling an LLM-based subtask scheduler, a shared topological memory that turns each agent's exploration into team knowledge, and a conflict-aware execution mechanism that realizes simultaneous intentions as collision-free routes. Extensive experiments establish TRISS as a comprehensive baseline and reveal substantial room for improvement across scheduling, planning, and execution, highlighting the challenges of coordinating under MAVLN task constraints. Project page: https://xyz9911.github.io/mavln. |
38 pa...38 pages, 18 figures, 16 tables |
| AirGroundVLN: A Large-Scale Benchmark for Goal-Oriented Air-Ground Collaborative Vision-and-Language Navigation | 2026-10-07 | ShowGoal-oriented Vision-and-Language Navigation (VLN) requires agents to locate and reach targets described in natural language without prescribed routes. Air--ground collaboration is valuable for tasks requiring both wide-area search and fine-grained localization. However, systematic study of goal-oriented air--ground collaborative VLN remains limited by the lack of large-scale, diverse benchmarks and two core challenges: 1) substantial differences between aerial and ground views, together with useful observations becoming unavailable as navigation proceeds, make it difficult to maintain spatially consistent context across platforms and over time; and 2) asymmetric spatial observability makes ground perception locally detailed but spatially limited and aerial perception broad but locally coarse, limiting the reliability of single-platform planning. To address these limitations, we introduce AirGroundVLN, a benchmark containing 10,281 navigation episodes and 955 target instances across 19 Unreal Engine environments, with seen/unseen splits and an aerial-visibility protocol for systematic evaluation. Alongside the benchmark, we propose AG-CoNAV, a trainable reference framework comprising two key components: Spatiotemporally Anchored Collaborative Memory (SACM) and Aerial-Guided Regional-to-Local Planning (AGRLP). SACM maintains and retrieves spatially consistent historical context across aerial and ground observations. Meanwhile, AGRLP combines regional aerial guidance with fine-grained ground navigation. Extensive experiments demonstrate the effectiveness of AG-CoNAV and establish AirGroundVLN as a comprehensive benchmark for future exploration. |
|
| Speaking the Navigator's Language: Trajectory-Grounded Instruction Translation for Frozen Aerial VLN Agents | 2026-10-07 | ShowAerial vision-and-language navigation (VLN) agents are typically trained on detail-rich, trajectory-aligned commands, whereas users issue short, intent-driven instructions; on a frozen OpenFly navigator, this \emph{instruction gap} drops success rate (SR) from |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| PolicyIK: High-Precision Inverse Kinematics for Robot Arms with Reinforcement Learning | 2026-11-03 | ShowInverse kinematics (IK) converts end-effector pose commands into joint motion, and many vision-language-action (VLA) policies rely on it to act on a given arm and to transfer across embodiments. Precise IK remains laborious: redundancy and singularities complicate the inverse mapping, execution depends on hand-tuned controllers, and training data on physical robots are costly to collect. We present PolicyIK, a goal-conditioned reinforcement learning policy for high-precision IK. It updates joint-position references from pose feedback and observes its accumulated reference relative to the measured joints, so it can retain the offset that the position controller needs at rest. Two-scale pose errors with an error-dependent action gain resolve millimeter residuals while preserving large motions. Progress and dwell rewards teach it to both reach the tolerance and stay within it. In simulation, a single attempt by the trained PPO policy keeps the mean errors over the final second within 1 mm and 0.2° on 93.5% of held-out tasks, and a deployment protocol raises success to 98.9%. When trained directly on a physical arm with an off-policy RL learner, PolicyIK achieves higher reaching and holding success than all four classical IK controllers under the same time budgets. In simulated pick-and-place, PolicyIK executes the end-effector commands of the |
|
| Fine-Tuning VLAs with Self-Demonstrated Generative Control for Multi-Task Manipulation | 2026-11-03 | ShowState-of-the-art vision-language-action (VLA) models such as pi0.5 exhibit strong semantic understanding, instruction following and task behavior. However, when deployed on new robots, even minor mismatches in hardware configuration relative to pretraining can cause severe performance drops. Finetuning the VLA on in-domain expert data from the new embodiment improves performance on the expert task but leads to a loss in its original instruction following and behavioral priors. In this paper, we propose a self-supervised method that generates online interaction rollouts from the zero-shot VLA as additional training data for finetuning. Our experiments show this yields strong multi-task manipulation policies on the target robot that (1) inherit pretrained tasks distilled from the zero-shot VLA without expert demos for them, (2) improve success and generalist instruction following beyond expert-only fine-tuning on expert-teleoperated tasks, and (3) mitigate forgetting of held-out tasks from the pretrained policy, not part of fine-tuning. We demonstrate the success of our approach across test sets probing generalization on a real ALOHA robot and a new simulation benchmark in RoboTwin. Video results are available at https://self-supervised-control.pages.dev/ |
Addit...Additional experiments. Project page, code, checkpoints, data: https://self-supervised-control.pages.dev/ |
| Pretraining Knows What, Simulation Teaches How: Compositional Generalization for Agentic Robotics | 2026-11-03 | ShowFrontier models such as GPT-6 Astra now directly control robots and can generalize better than vision-language-action models (VLAs). We argue that this ability could arise from two distinct, composable skills: semantic reasoning, to understand what a task asks for in a scene, and motion control, to move the robot to do it. We show that small open VLMs are already capable of the first and mainly lack the second. We introduce Zero-real Agent Policy (ZAP), a recipe that teaches this missing motion skill, entirely in simulation on simple primitive tasks and from the model's own successes, with no frontier teacher or distillation and no real-world robot data. ZAP reaches 96.7% success on real-world tasks, matching GPT-6 Astra in our harness while running 3x faster and 17x cheaper, and 74.0% on a bimanual adaptation of RoboLab-120, on par with GPT-6 Astra's 72.2%. ZAP also generalizes to new objects, goals, and strategies, and in simulation to a robot embodiment on which it never trained. |
|
| Should the Action Expert Train the VLM? Action-Token Supervision and Gradient Routing in Mixture-of-Transformer VLAs | 2026-11-03 | ShowVision-language-action (VLA) models increasingly pair a pre-trained vision-language model (VLM) with a flow-matching expert that conditions its action predictions on the VLM's key-value representations. Those representations must carry information useful for control, yet the VLM was never trained for the action expert's flow-matching objective. Two signals can address this: gradients backpropagated from the expert's loss (coupling), or VLM supervision to predict discrete tokens encoding the same target actions. Prominent training recipes forgo the first, blocking expert-to-VLM gradients (insulation) to protect the VLM's existing capabilities, while also supervising it on vision-language data and action tokens. Because vision-language co-training also aims to preserve those capabilities, the additional protection provided by insulation is unclear. At the same time, insulation removes a direct signal about which features continuous action prediction requires. On the other hand, action-token supervision offers an alternative way to learn action-relevant representations, which may explain why insulation succeeds in practice. We disentangle these factors in autonomous driving and manipulation by independently varying action-token supervision and whether the expert's loss updates the VLM. In open-loop driving, insulation is competitive with coupling when action tokens are supervised; without them, coupling reduces trajectory-prediction error by roughly 25%. When both components are trained jointly from the outset, coupling increases closed-loop driving route progress by approximately 17% with action-token supervision and 40% without it, and similarly improves robot manipulation performance under both settings. These findings show that the value of gradient insulation depends on the VLM's action supervision and the evaluation setting, challenging insulation as a default VLA training recipe. |
|
| Asynchronous Robot Policies | 2026-11-03 | ShowScaling vision-language foundation models has significantly advanced generalist robot manipulation, but the increased model scale introduces computational latency that degrades real-time physical reactivity. Existing Vision-Language-Action (VLA) architectures execute high-level semantic reasoning and action ex- ecution sequentially within the primary control loop, creating a fundamental tension between semantic-reasoning capability and robot agility. To address this, we propose Asynchronous Robot Policy (ARP), an architectural framework that decouples semantic reasoning from low-level action generation using asynchronous processes operating at independent frequencies. In ARP, a high-capacity Vision-Language Model (VLM) asynchronously updates a semantic memory at low fre- quency, while a lightweight Vision-Action Expert, conditioned on this semantic memory, generates robot actions at a much higher control frequency. Such a formulation, enables having a fast perception-action loop without compromising on the semantic capability. To overcome asynchronous memory staleness, ARP trains the vision-action expert on the semantic staleness and execution delay, with temporally mismatched observation across processes. On Franka manipulation tasks, ARP achieves an 18% higher success rate and 20% less time need for execution in quasi-static setups, while 3-5 times baseline performance in dynamic simulation environments. Furthermore, ARP enables operating the VLM at 10 times lower than its nominal frequency with no significant performance loss making it much more efficient than the baseline. These results highlight the architectural advantages of modeling robot policies as asynchronous processes, enabling real-world reactive deployment without sacrificing the scale of foundation models. |
For c...For code, checkpoints and more results, check out https://arp.insait.ai |
| LoopQ: Fair, Closed-Loop Evaluation of Quantized Vision-Language-Action Models | 2026-11-03 | ShowPost-training quantization (PTQ) for vision-language-action (VLA) models promises smaller and faster VLA deployments. However, when analyzing existing VLA quantization works, we realised that it is hard to reason about the performance of the suggested quantization method compared to existing, similar, or in many cases, simpler PTQ methods. We present LoopQ, a closed-loop benchmark that compares multiple quantization policies on the same checkpoint, layers and engine, varying only the method and the kernels. We evaluate RTN, AWQ, GPTQ, ActQuant and QuantVLA on four VLAs with more than 180,000 episodes of LIBERO and LIBERO-plus, compare against QVLA's reported results on the same layers, and benchmark whether the four public VLA-quantization releases can be reproduced. We benchmark the methods on an NVIDIA H100, an L4, and a Jetson AGX Thor. Our results show that while quantization reliably saves memory, 2-3x for real low-bit models, VLA-specific PTQ methods underperform compared to simpler generic baselines. At 4 bits, RTN, AWQ and GPTQ match or exceed ActQuant and QuantVLA, plain RTN matches the calibrated methods, and plain AWQ matches the success QVLA reports for itself. Standard LIBERO hides the losses that do exist: configurations with no significant LIBERO loss lose 1.5-4.3 points under paired LIBERO-plus perturbations. Speed is a weaker story. At batch 1 on the H100, no quantized pi0.5 or OpenVLA-OFT configuration is faster than its 16-bit model (1.06-1.58x slower on library kernels); only the L4 with int8 kernels gains (4-14%), the Thor is slower on every 4-bit configuration, and the engine alone changes latency by 2.1x. Profiling explains why: pi0.5's quantized backbone is compute-bound, while its memory-bound action expert, 63% of GPU time and 13% of the parameters, stays in 16 bits. |
|
| CRONOS: Benchmarking Non-Episodic Reinforcement Learning for Robotic Manipulation | 2026-11-03 | ShowReinforcement learning (RL) is promising for adapting robot policies to unstructured real-world environments. However, standard RL pipelines rely on episodic training with frequent scene resets, which is impractical for real-world deployment due to the need for substantial human intervention. While prior work has studied non-episodic RL in relatively small policies trained from scratch, its behavior when fine-tuning modern large-scale pre-trained vision-language-action (VLA) models remains underexplored. In this work, we introduce CRONOS, a simulation benchmark for RL fine-tuning of VLAs under non-episodic scene resets. CRONOS leverages high-fidelity physics simulators and adopts shared-scene multi-task settings. Using CRONOS, we show that naively fine-tuning pre-trained policies fails in non-episodic settings due to shifts in object configurations and end-effector poses; however, these challenges can be mitigated through targeted reset allocation and by addressing biases in pre-trained models. Moreover, we find that non-episodic training enhances long-horizon manipulation and improves generalization to previously unseen object configurations and task sequences. Project page: https://embodiedai-ntu.github.io/cronos/ |
|
| Caging Vision-Language Navigation for Legged Robots via Barrier-Certified Safety Architecture | 2026-11-03 | ShowBy harnessing the inherent commonsense capabilities of pretrained Large Language Models (LLMs) and Vision-Language Models (VLMs), current Vision-Language Action(VLA) models for legged robots excel at interpreting language-based navigation commands but lack the physical grounding and explicit safety assurance required for real-world operation. Therefore, this paper aims to cage Legged-Vision-Language Navigation (VLN) through a safety architecture that provides physical grounding and ensures navigation safety. The key idea is not to abandon the powerful capabilities of end-to-end VLA planning, but rather to augment and cage them. In particular, we integrate a high-level VLA planner with a low-level physical-grounding layer that leverages efficient state estimation, in-situ 3D spatial perception, and an Signed Distance Field (SDF) representation to generate executable trajectories. We provide a navigation planner with provable collision-avoidance guarantees under the stated assumptions by combining a low-dimensional Bezier-parameterized MPPI planner with Discrete Exponential Barrier Function (dEBF) constraints. The proposed Legged-VLN system is modular, making it compatible with various VLM/VLA models and agnostic to the robot platform. The proposed system is validated with extensive real-world experiments on humanoid and quadrupedal robots, demonstrating versatile, efficient, and safe vision-language driven navigation capabilities. |
|
| Trusting Imagined Futures: Auditing Imagined Rewards for VLA Post-Training | 2026-11-03 | ShowVideo world models enable candidate actions to be evaluated through imagined outcomes, but high predicted rewards can favor trajectories with poor execution quality. We introduce Trusting Imagined Futures (TIF), an offline supervision selection method for vision-language-action (VLA) post-training that audits predicted rewards using execution feedback. A multidimensional reward model predicts a utility score for candidate ranking and five dimensions of execution quality. An auditor uses reward predictions and uncertainty to identify overestimation patterns learned from paired imagined videos and real executions. TIF accepts the candidate with the highest predicted utility when its advantage over the existing trajectory exceeds a margin and its audit risk passes a threshold; otherwise, it retains the existing supervision. The retained real trajectories then supervise behavior cloning. Across four LIBERO suites and three training seeds, TIF achieves a mean success rate of 76.25%, improving over training on the existing trajectories by 3.00 percentage points with the same trajectory count and training budget. Adding audit gating yields 1.47 percentage points with the reward model and advantage margin held fixed. The gains persist against advantage ranking and logistic regression risk selection at matched replacement counts. On recorded videos from three real robot tasks, TIF increases clean completion among selected trajectories from 58.33% to 70.00% across 60 candidate pools. |
|
| LIBERO-RECOVER: Beyond Task Success Towards Failure Recovery in Robotic Manipulation Models | 2026-11-03 | ShowVision-Language-Action (VLA) or World Action (WAM) models have recently demonstrated remarkable performance in robotic manipulation. On LIBERO, SOTA method have achieved nearly 100% success rates, seemingly suggesting that the models are ready for deployment in real world. However, near perfect performance on existing benchmarks can be misleading: success under ideal conditions does not imply real world robustness. Existing benchmarks primarily evaluate task completion from predefined initial states, while real world interactions inevitably involve failures such as failed grasps, collisions, and unintended object movements. A robot must therefore not only execute tasks successfully, but also recognize and recover from failures to continue the task. Yet this capability remains largely unmeasured, revealing a critical gap between benchmark performance and real world reliability. To address this gap, we introduce LIBERO-Recover Benchmark, a large scale benchmark for failure recovery in robotic manipulation. Built upon LIBERO, we collect real execution failures from SOTA embodied models and construct 1,000+ scenarios across four recovery levels: (1) Action Retry, (2) Action Adaptation, (3) Object State Recovery, and (4) Environmental Recovery. We evaluate four core capabilities: spatial understanding, object structure reasoning, interaction understanding, and topological reasoning. As the first large-scale benchmark for embodied failure recovery, LIBERO-Recover shifts evaluation from \emph{Can the robot succeed?''} to \emph{Can the robot recover after failure?''}, promoting robust and generalizable embodied agents. The project will be avaible in \textcolor{blue}{https://liulin815.github.io/LIBERO-Recovery/}. |
|
| PRIME W0: Physical Priors from World Models for Embodied Manipulation | 2026-11-03 | ShowVision-Language-Action (VLA) models excel at instruction following, yet most remain feed-forward: they condition actions only on the current observation and discard both long-horizon temporal context and the hierarchical structure of VLM representations. We present PRIME W0, a hierarchical conditional flow-matching framework that closes this gap through three synergistic designs: (1)per-layer KV injection from a pretrained VLM into an Adaptive MoE action expert; (2)dynamics-aware conditioning via asynchronous world-model subgoal-image prediction and a DINOv2 historical-frame semantic bridge; and (3)Action Dynamic Manifold (ADM) flow matching with covariance-structured noise that respects robot kinematic coupling. PRIME W0 is trained in four stages---VLM pre-training, DAT/NTP visuomotor pre-training, DiT flow-matching decoder assembly with WM injection, and end-to-end SFT---and achieves state-of-the-art results on LIBERO and real-robot benchmarks, with systematic ablations validating each architectural component. |
|
| CLAP: Direct VLM-to-VLA Adaptation via Language-Action Grounding | 2026-11-03 | ShowVision-language-action models (VLAs) inherit semantic capabilities from pretrained VLMs, yet large-scale post-training on robot data and architectural modifications can reshape the backbone so extensively that it becomes difficult to isolate what the VLM contributes to control. Directly converting pretrained VLMs into VLAs with minimal architectural change offers a more transparent path to understanding how VLM capabilities transfer across model scales. The core obstacle is output-distribution mismatch: predicting actions as bare numeric token sequences moves generation away from the VLM's pretrained language distribution, degrading the capabilities we seek to preserve. To address this, we propose CLAP (Causal Language-Action Prediction), which prepends each numeric action sequence with a natural-language action description, causally conditioning precise action-token prediction on a language-action plan without modifying the backbone architecture. With only 4-5 hours of training, 2B CLAP achieves 90.8% on LIBERO (+14.9 pt over VLA-0) and improves robustness on LIBERO-PRO under language, object, and spatial perturbations. Ablations attribute the gains to action-grounded language semantics, and on a real UR5e robot 2B CLAP reaches 60% success against 15% for matched VLA-0. We release CLAP at 0.8B, 2B, and 4B as an open-weight, multi-scale compact VLA family from a single VLM lineage, enabling controlled analysis of VLM-to-VLA capability transfer. |
Accep...Accepted to CoRL 2026. Project page: https://omron-sinicx.github.io/clap |
| What the Guard Misses, the Robot Executes: Implied Harm in VLA Instructions | 2026-11-03 | ShowVision-language-action models (VLAs) act on instructions without being able to refuse, so screening harmful requests falls to monitors. We test whether these monitors catch ordinary robot tasks requested for harmful reasons, holding the task fixed while varying only how explicitly the intent is stated. |
Accep...Accepted at the Science of Physical AI Safety (SPAIS) Workshop at CoRL 2026 |
| Action Intent as a Bridge: Transferring Manipulation Knowledge from Human-Centric Videos to Robots | 2026-11-03 | ShowVision-Language-Action (VLA) models have achieved promising performance for generalized robotic manipulation, yet their reliance on expensive real-robot data remains a fundamental bottleneck. A scalable alternative is to transfer knowledge from off-the-shelf human-centric videos. However, existing methods extract latent action representations that are entangled with scene-specific noise, diluting the transferred knowledge. Motivated by this limitation, we propose Action-Intent Transfer (AIT), a two-stage method that transfers knowledge through a clean action representation shared across humans and robots. In the first stage, AIT extracts explicit action intents and disentangles the implicit action intent from manipulation videos. In the second stage, AIT aligns the human and robot domains within this shared intent space and decodes the intent into executable robot actions. Extensive experiments in both simulation and real-world environments demonstrate that AIT transfers human knowledge more effectively and consistently outperforms prior baselines. |
12 pages,6 figures |
| Aero-VPA: Few-Step Flow-Matching Control over Non-Reconstructive JEPA Latents for Agile Flight on Edge Hardware | 2026-11-03 | ShowVision-Language-Action (VLA) models generalize across tasks, but their autoregressive decoders need sequential forward passes in proportion to the action horizon, each bounded below on embedded aerial platforms by streaming billions of parameters from memory. We present Aero-VPA, a lightweight Vision-Predictor-Action architecture for 3D navigation on edge hardware. A compact RGB-D Vision Transformer and a five-head ensemble predictor, trained without pixel reconstruction, forecast the next latent state; a Conditional Flow Matching solver generates a velocity and yaw-rate chunk in two Euler steps, so each decision takes five sequential network evaluations; and the ensemble dispersion drives an uncertainty fallback. Version 1 specified the architecture, deployment and protocol without results. This version reports the first closed-loop evaluation: PX4 software-in-the-loop simulation of quadrotor missions in procedurally generated forests. Behaviour cloning failed on all 60 missions despite open-loop errors of 0.02-0.04 m/s: the policy copies its last command, a copycat form of causal confusion. DAgger with a privileged, safety-gated supervisor gave the first successes but plateaued at 11.1-21.1% across six checkpoints. A map-free runtime shield with a 15 s depth memory, barrier constraints, a cap on motion outside the 73-degree field of view and backoff along the flown path raised the first DAgger policy's success on 180 paired missions from 21.1% to 52.2% (95% CI [45.0, 59.4]; exact McNemar p = 1.2 x 10^-10) and cut its crashes from 72.2% to 22.2%, while timeouts rose from 2.8% to 23.3%. All six checkpoints gained 30.0-34.4 percentage points, at mean speeds of 1.16-1.36 m/s. The code (https://github.com/ylee-sling/Aero-VPA), the trained policy (72.6 MB, https://huggingface.co/ylee-sling/aero-vpa) and the full benchmark dataset are public. |
38 pa...38 pages, 3 figures, 22 tables. Simulation study (PX4 SITL + Gazebo Harmonic). Version 2 of a preprint first posted on Zenodo (doi:10.5281/zenodo.23633375). Code: https://github.com/ylee-sling/Aero-VPA ; trained policy: https://huggingface.co/ylee-sling/aero-vpa ; benchmark dataset: https://huggingface.co/datasets/ylee-sling/aero-vpa-benchmark |
| Human-to-Robot Transfer via Bridging Representation and Action Space in VLAs | 2026-11-03 | ShowPretrained Vision-Language-Action (VLA) policies provide a strong starting point for deployment, but adapting them to a target robot and environment still requires costly embodiment-specific demonstrations. Human demonstrations are far easier to collect, yet directly leveraging them is challenging due to the substantial human-robot embodiment gap, particularly when the underlying VLA has been pretrained primarily on robot data. We introduce H2RB (Human-To-Robot Bridge), a human-robot co-training framework for adapting pretrained VLA policies from unpaired human and robot demonstrations. H2RB bridges the embodiment gap at both interfaces: geometry-anchored alignment grounds human and robot observations in a compatible representation, while a decoupled-decoder tokenizer shares action latents without forcing both domains through a single action mapping. This design allows human demonstrations to provide useful behavioral supervision while retaining the rich visual, semantic, and control priors of the pretrained VLA. Across real-world tasks, H2RB substantially improves object transfer performance from 8.0 to 51.8 and enables skill transfer to score 62.7, in settings where co-train baselines totally fail to transfer human-demonstrated skills. The benefit also persists under object-placement shifts, improving out-of-distribution transfer from 6.0 to 45.3. H2RB suggests a way for human demonstrations to serve as a scalable supervision source for adapting pretrained robotic policies, reducing the dependence on costly target-robot data collection. |
Accep...Accepted for Oral Presentation at the CoRL 2026 Workshop on Scaling H2R. Project page: https://junhyeong-park.me/h2rb/ |
| RAFT: Reasoning-Aligned Fine-Tuning Improves Robustness of VLA Post-Training | 2026-11-03 | ShowAdapting a generalist vision-language-action (VLA) policy to a target task with limited demonstrations can improve nominal task success while reducing robustness to variations in appearance, objects, and instructions. We introduce reasoning-aligned fine-tuning (RAFT), a two-stage post-training procedure: RAFT first trains low-rank adapters on task-relevant reasoning annotations, then freezes them while fine-tuning the policy on action demonstrations. On a physical DROID robot, RAFT achieves |
|
| AceAction-5M: Hierarchical Goals for Generalist Action Learning Across 3D Game Worlds | 2026-11-03 | ShowVideo games provide a practical setting for collecting vision-language-action (VLA) data, whose acquisition in physical environments is costly. Existing datasets, however, remain limited by restricted game coverage, sparse language supervision, or temporal hierarchies that do not necessarily reflect changes in task objectives. We introduce AceAction, a dataset of human gameplay across multiple 3D games for learning skill composition and subgoal transitions. AceAction organizes demonstrations around complete tasks, linking overall instructions, ordered subgoals, and local skill instructions to temporally aligned visual observations and keyboard-and-mouse actions. To construct these annotations, we develop a vision-language model pipeline that identifies subgoals from changes in objectives or interaction targets, anchors navigation goals to visually recognizable waypoints, and composes adjacent subgoals into task instructions. The resulting demonstrations cover navigation, reconnaissance, combat, interaction, and recovery, with explicit stage boundaries that connect task progression to fine-grained control. We assess annotation quality through model-based verification and human inspection. An audit of 500 trajectories yields a 95.8% pass rate across independent reviewer assessments. AceAction provides a data foundation for studying how agents execute, combine, and switch skills to follow complex instructions in dynamic game environments. |
|
| DiMuRe: Diagnosis-Guided Multilevel Recovery for Robotic Task Execution | 2026-11-03 | ShowDisturbances can invalidate the conditions required by a robot's next action even when much of its task program remains useful. A recovery system must therefore choose a response suited to the current state. We present DiMuRe (diagnosis-guided multilevel recovery), a framework for robot-use agents that uses phase-specific evidence and method preconditions to select program reuse or predefined recovery skills, including bounded vision-language-action (VLA) grasping. After the original program exits, a multimodal large language model (MLLM) can generate corrective code from images and execution feedback. Local verification checks the outcome of each recovery attempt. We also introduce LIBERO-Disturb, a benchmark for robustness and recovery under mid-execution disturbances, supporting program-based agents and VLA policies. On 801 scheduled disturbed Encore development cases, Local DiMuRe (program reuse and recovery skills) achieves 77.65% complete-task success, compared with 58.68% for the original program under the same physical-step allowance. Full DiMuRe achieves 81.90% with terminal replanning, an additional step allowance, and early handoff. Local DiMuRe also uses fewer physical steps per scheduled episode than the original program. |
15 pa...15 pages, 3 figures, including appendices |
| Say the Same, Act Differently: Diagnosing and Steering Text-Action Divergence in Reasoning VLAs | 2026-11-03 | ShowReasoning vision-language-action (VLA) models pair physical control with natural-language rationales that supervisors can read to monitor embodied safety. Prior attacks on VLA reasoning corrupt the rationale itself and leave visible evidence in the text. We show that visual perturbations can instead alter the physical action while keeping the generated rationale word-for-word identical. At the shared text/action interface of Alpamayo for driving and InstructVLA for manipulation, our text-neutral action-subspace (TNAS) diagnostic shows that the locally most action-sensitive direction is nearly orthogonal to the active text-loss gradients, with 1.87% and 5.28% norm overlap. At equal hidden-state perturbation norms across both models, interventions along this text-neutral residual shift actions 4.64x to 47.8x more than text-gradient perturbations, whereas text-gradient perturbations induce 21.2x to 277x more text drift. Steering visual perturbations along this direction reaches the same decoupling from the image. Across 154 obstacle-annotated driving scenes in Alpamayo, TNAS-guided attacks induce a new collision under a word-for-word identical rationale in 52.6% of scenes under full-image perturbations and 35.1% under a localized forward-camera patch, compared with 4.55% for unguided action attacks. Relative to a text-preserving baseline, TNAS guidance raises exact rationale preservation from 83.0% to 89.5% and cuts teacher-forced text drift by 68.4% at a comparable collision rate. In InstructVLA, TNAS guidance cuts text drift by 40.5% and preserves the exact response on 85.0% of examples alongside a 1.73-unit action-chunk shift. An unchanged rationale therefore cannot certify a safe action, so monitors for reasoning VLAs must verify the predicted action channel directly. |
Accep...Accepted to NeurIPS 2026. 23 pages, 3 figures, 8 tables |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| YANAMI: A Large-Scale Multimodal Dataset and Benchmark for Multiturn Stateful Interaction | 2026-11-03 | ShowUnified Multimodal Models (UMMs) integrate visual understanding, generation, and editing within a single unified framework. However, existing unified multimodal datasets and benchmarks remain dominated by single-turn tasks. Even in recent multi-turn datasets, many turns can be solved independently without tracking information across turns. Consequently, they provide limited supervision and evaluation for multimodal stateful interaction, which requires models to understand, maintain, update, and faithfully realize evolving dialogue states across turns. To address this gap, we introduce YANAMI, the first large-scale multimodal dialogue dataset and benchmark for stateful interaction, comprising YANAMI-400K and YANAMI-Bench. YANAMI-400K contains 400K multi-turn multimodal dialogues, with more than 2M images and 2.6M dialogue turns. The dialogues span four representative scenarios: Multi-Image Fusion, Image Editing, World Modeling, and Visual Reasoning, and require models to continuously track, preserve, and revise evolving multimodal dialogue states across turns. YANAMI-Bench comprises 400 multi-turn dialogues for systematically evaluating multimodal stateful interaction. It jointly evaluates Dialogue State Understanding and Visual Realization, characterizing the gap between models' ability to understand evolving dialogue states and their ability to faithfully realize them in visual outputs. Experiments show that fine-tuning BAGEL on YANAMI-400K improves image editing, generation, and multimodal stateful interaction by 10.3%, 2.0%, and 13.23%, respectively, while preserving its general visual understanding capabilities. Evaluations on YANAMI-Bench further reveal that existing UMMs still struggle to maintain, update, and visually realize evolving multimodal dialogue states. |
Proje...Project page: https://rowletqwq.github.io/YANAMI |
| Trajectory Evaluation via Rollout-Free World Model for End-to-end Autonomous Driving | 2026-11-03 | ShowAlthough existing candidate trajectory evaluators have substantially improved end-to-end planning accuracy, their underlying mechanism remains suboptimal. Model-based evaluators typically make scoring future-aware by explicitly predicting candidate-conditioned future states over the planning horizon, which we term rollout. However, rollout incurs high inference cost and accumulates prediction errors. We argue that planning does not require reconstructing a full future scene for each candidate, but only future-relevant information for reliable comparison. We therefore propose RFWorld, a Rollout-Free World model for end-to-end trajectory evaluation. RFWorld constructs a shared scene memory and shapes it into a future-readable representation via lightweight training-time temporal readout with semantic supervision, keeping training overhead low. At inference time, candidate trajectories directly query the shared memory via sparse spatial readout for scoring, reducing computation by roughly the planning horizon times the number of candidate trajectories, while avoiding recursive rollout errors. Experiments on NAVSIM and Bench2Drive show that RFWorld learns a more planning-aligned world representation and achieves state-of-the-art performance. |
|
| World-Model-Augmented Visual Locomotion for Humanoids on Foothold-Constrained Terrain | 2026-11-03 | ShowFoothold-constrained terrain limits feasible footholds to sparse, discontinuous, or narrow support regions, as in gaps, stepping stones, stairs, and balance beams. In such settings, a single misplaced step can cause a fall. Reliable traversal therefore requires precise foot placement and anticipation of upcoming terrain. To address this challenge, we present World-Model-Augmented Visual Locomotion (WM-LOCO), which augments a visual locomotion policy with predictive context from a jointly trained world model. Given onboard depth observations and proprioception, a recurrent state-space model (RSSM) learns action-conditioned latent dynamics that predict future proprioceptive states, egocentric observations, and task rewards. The resulting recurrent feature is then fed into the proximal policy optimization (PPO) actor-critic, providing predictive context for anticipatory foot placement on foothold-constrained terrain. WM-LOCO achieves average success rates of 96.0% on gaps, 96.9% on stepping stones, 94.0% on stairs, and 92.1% on balance beams, outperforming the evaluated baselines while improving gait efficiency and smoothness. Furthermore, WM-LOCO achieves zero-shot sim-to-real transfer to a Unitree G1, with 92.5% success across these four terrain classes. |
8 pag...8 pages, 6 figures, 4 tables. Yuxi Liu and Lijun Han contributed equally |
| Trusting Imagined Futures: Auditing Imagined Rewards for VLA Post-Training | 2026-11-03 | ShowVideo world models enable candidate actions to be evaluated through imagined outcomes, but high predicted rewards can favor trajectories with poor execution quality. We introduce Trusting Imagined Futures (TIF), an offline supervision selection method for vision-language-action (VLA) post-training that audits predicted rewards using execution feedback. A multidimensional reward model predicts a utility score for candidate ranking and five dimensions of execution quality. An auditor uses reward predictions and uncertainty to identify overestimation patterns learned from paired imagined videos and real executions. TIF accepts the candidate with the highest predicted utility when its advantage over the existing trajectory exceeds a margin and its audit risk passes a threshold; otherwise, it retains the existing supervision. The retained real trajectories then supervise behavior cloning. Across four LIBERO suites and three training seeds, TIF achieves a mean success rate of 76.25%, improving over training on the existing trajectories by 3.00 percentage points with the same trajectory count and training budget. Adding audit gating yields 1.47 percentage points with the reward model and advantage margin held fixed. The gains persist against advantage ranking and logistic regression risk selection at matched replacement counts. On recorded videos from three real robot tasks, TIF increases clean completion among selected trajectories from 58.33% to 70.00% across 60 candidate pools. |
|
| PrimeSim: A Generalist World Model Simulator from Large-Scale Cross-Embodiment Demonstrations | 2026-11-03 | ShowA robot world model that serves as a simulator must do more than generate plausible videos: it must faithfully follow commanded actions, maintain consistency across camera views, and sustain coherent long-horizon rollouts. We present PrimeSim, an action-conditioned, multiview world model simulator that addresses these challenges through progressive midtraining and posttraining. Starting from a pretrained video generator, we midtrain on 30,110 hours of robot demonstrations spanning 22 platforms. To improve action following, we introduce a unified cross-embodiment action representation and frame-aligned dual-path conditioning to strengthen action-frame correspondence. To maintain multiview consistency, we develop view-aware interleaved attention with explicit camera identities, enabling cross-view information exchange while preserving view-specific representations. We further introduce a fast-video, slow-memory posttraining framework that transforms the bidirectional video generator into an efficient autoregressive simulator. By combining fast causal generation, self-rollout training, and hierarchical temporal memory, PrimeSim mitigates error accumulation from self-generated histories and supports coherent long-horizon simulation. PrimeSim runs at 10.1 FPS, remains stable over long-horizon rollouts on rigid and deformable manipulation tasks, and transfers to held-out platforms within 100 to 200 gradient steps where the pretrained backbone does not. Against five open-source action-conditioned world models it leads four of the five WorldSimProbe suites and posts the best overall EWMBench score, while fine-grained local calibration still trails the field; its rollouts are faithful enough to support GRPO-based policy optimization. We release the model weights, inference code, and training code. |
|
| PRIME W0: Physical Priors from World Models for Embodied Manipulation | 2026-11-03 | ShowVision-Language-Action (VLA) models excel at instruction following, yet most remain feed-forward: they condition actions only on the current observation and discard both long-horizon temporal context and the hierarchical structure of VLM representations. We present PRIME W0, a hierarchical conditional flow-matching framework that closes this gap through three synergistic designs: (1)per-layer KV injection from a pretrained VLM into an Adaptive MoE action expert; (2)dynamics-aware conditioning via asynchronous world-model subgoal-image prediction and a DINOv2 historical-frame semantic bridge; and (3)Action Dynamic Manifold (ADM) flow matching with covariance-structured noise that respects robot kinematic coupling. PRIME W0 is trained in four stages---VLM pre-training, DAT/NTP visuomotor pre-training, DiT flow-matching decoder assembly with WM injection, and end-to-end SFT---and achieves state-of-the-art results on LIBERO and real-robot benchmarks, with systematic ablations validating each architectural component. |
|
| A research world model of power engineering learns its dynamics and reveals what-ifs | 2026-11-03 | ShowDecisions about the direction of research in power engineering determine which technologies mature and, through them, shape infrastructure that lasts decades, yet their consequences cannot be observed before they are taken, and existing forecasts of research fields extrapolate a single trajectory. Here we introduce research world models, which treat a research field as a dynamical system that can be acted upon, and establish one for power engineering from 1.1 million papers. A distilled language model reads each paper into role-typed concepts, two sophisticated, high-fidelity graph-temporal neural network cores learn how the resulting concept network evolves, and operators learned from natural experiments in the publication record estimate the effects of interventions. In temporally held-out tests, the model forecast years ahead which concepts would grow and which method-target pairs would be joined for the first time, outperforming tuned machine-learning baselines in most settings, and its estimated intervention effects tracked those realized in natural experiments it had not seen. Its training scheme and three of its operators replicated in two fields withheld from development. Questions posed in natural language are compiled by a domain-specialized 27-billion-parameter language model into executable actions and answered as paired simulations with and without the intervention. The answers were rated 4.1 of 5 for faithfulness by an independent language-model judge. By revealing the latent dynamics of a research field and its response to interventions, research world models offer a new way to simulate and test research decisions before they are made. |
|
| Hybrid Particle-Voxel Dynamics for Multimodal 3D World Simulation | 2026-11-03 | ShowPredicting how the world evolves in response to robot actions is central to policy learning, evaluation, and planning. However, many action-conditioned world models predict future observations in 2D image or video space without explicitly modeling the underlying 3D scene, making spatial consistency across viewpoints and over long-horizon interactions challenging. We introduce 3D World Simulator, an action-conditioned multimodal world modeling framework that jointly predicts appearance, geometry, and tactile observations from a unified 3D representation. Our approach combines a hybrid particle-voxel autoencoder that compresses point clouds into compact 3D voxel latents with an action-conditioned latent dynamics model that predicts their evolution. The shared 3D representation supports geometrically consistent rendering from arbitrary viewpoints and flexible integration of spatially aligned tactile signals. Our experiments demonstrate interactive, long-horizon, multimodal prediction at 50 Hz. We then show how our model enables scalable data generation for policy training and evaluation using tactile and multi-view visual observations, including wrist-mounted and novel camera views. Furthermore, we scale our method to large datasets containing diverse objects and demonstrate model-based planning with unseen objects and robot configurations. Together, these results highlight the potential of hybrid particle-voxel dynamics as a foundation for multimodal 3D world simulation. |
|
| Embodied AI for Scientific Discovery: From Physical Action to Scientific Evidence | 2026-11-03 | ShowScientific experiments are increasingly selected, executed and interpreted by learned systems that act through physical interfaces. Robot learning builds physical competence, whereas AI for science chooses which experiment to run, each reporting different endpoints. This review treats their interface as one problem, embodied AI for science, where the experiment itself is partially observed: specimen state, apparatus state and measurement processes are latent, actions are costly and irreversible, and missingness depends on execution and outcome. We organize the literature by three closures operating at distinct levels: operational closure (the experiment realized is the one intended), epistemic closure (the result updates beliefs or selects subsequent experiments) and evidential closure (the record supports a claim). We analyze workflows through a record-linked evidence cycle - Record, Diagnose, Calibrate, Confirm - as analytic functions with feedback. A taxonomy crosses these closures with six delegated decisions and five embodiment classes. Our formulation separates commanded, realized and observed conditions and models diagnosis with explicit resource, risk and sample-change costs. We compare experiment-selection methods by terminal decision, review perception, protocol compilation, visuomotor policies, world models, recovery and transfer by evidence emission requirements, and compare chemistry, materials, biology, microscopy and field workflows by what their records establish. An appendix provides an evidence map of 28 workflows (26 physical, two software-boundary), characterizing validation pathways and claim confirmation requirements. We propose evaluation that separates actions, valid experiments, campaigns and confirmed claims, framing open challenges as controlled comparisons with measurable failure conditions. |
46 pa...46 pages, 10 figures, 14 tables. Review article |
| Medical world models in healthcare: capabilities, evidence, and trustworthy clinical translation | 2026-11-03 | ShowMedical world models offer a framework for extending medical artificial intelligence beyond static prediction by representing evolving patient states and modelling how they change over time and in response to clinical interventions. This Review uses structured narrative synthesis and reproducible evidence mapping to define conceptual boundaries, technical foundations, application domains, and clinical evidence requirements. We screened 1,455 unique records and assembled a corpus of 300 sources, including 14 studies that met a strict empirical definition of a medical world model. The field is organized around four capabilities: patient state representation, temporal dynamics modelling, intervention-conditioned simulation, and clinician-supervised planning. Evidence spans medical imaging, longitudinal electronic health records, treatment response modelling, physiological and multimodal state modelling, ultrasound and surgical interaction, and population and health-system simulation; clinical digital twins are treated as a cross-cutting integration framework. Studies provide early evidence of technical feasibility for trajectory forecasting and comparison of candidate interventions, but most remain retrospective, task-specific, or preclinical. We distinguish functional capability from clinical evidence maturity and specify evidence needed to support progressively stronger clinical claims, from temporal forecasting to intervention comparison and planning-oriented decision support. These requirements include precise intervention representations, robust causal and mechanistic grounding, calibrated trajectory-level uncertainty, safety-constrained planning, external validation, and prospective multicentre evaluation against clinically meaningful endpoints. This framework aims to distinguish technically plausible simulation from clinically supportable prediction and decision support. |
Updat...Updated title, author order, and abstract; expanded the source corpus to 300 sources |
| Distill to Think, Foresee to Act: Cognitive-Physical Reinforcement Learning for Autonomous Driving | 2026-11-03 | ShowCurrent end-to-end autonomous driving models are fundamentally constrained by the behavioral cloning ceiling of imitation learning. While reinforcement learning offers a path to smarter autonomy, it demands two missing pieces of infrastructure: (1) a cognitive foundation that understands traffic semantics and driving intent, and (2) a foresighted physical environment that can anticipate the consequences of candidate actions. To this end, we propose CoPhy, a CognitivePhysical reinforcement learning framework for autonomous driving. To distill to think, we distill VLM knowledge into the BEV encoder and then discard the VLM entirely, retaining cognitive ability at zero inference cost while releasing the cognitive channel as a pluggable interface for optional human language commands. To foresee to act, we build an auto-regressive BEV world model that explicitly predicts future semantic maps conditioned on candidate actions, serving as an interpretable physical sandbox from which safety metrics are directly derived. Built upon this dual infrastructure, we optimize the driving policy via GRPO with a novel dual-reward mechanism: a physical reward derived from BEV rollouts enforces hard safety constraints, while a cognitive reward from a language-aligned scorer ensures intent compliance. Extensive experiments demonstrate that CoPhy not only achieves state-of-the-art results on NAVSIM v1 and v2 benchmarks, but also enables safer driving via cognitively informed scene compliance and flexible intent control through user-defined language instructions. Project page: https://wuyang98.github.io/CoPhy/. |
|
| See, Foresee, Act: Intent-Aware World Modeling for Proactive Retail Assistance | 2026-11-03 | ShowWorld models allow agents to anticipate action consequences, but proactive assistance also requires interpreting incomplete observations and deciding whether to intervene before a user asks for help. We propose See-Foresee-Act and instantiate it as the Proactive Intent World Model (PIWM) for retail assistance. See preserves role-relevant observable evidence alongside AIDA-BDI customer states; Foresee predicts one-step consequences for candidate actions; Act explicitly compares those consequences, including non-intervention. GuidanceSalesBench supplies predominantly synthetic supervision linking observations, states, candidate outcomes, and decisions. On 509 Full test examples, consequence-based selection achieves 63.65% decision accuracy, compared with 45.38% for direct decision output, an improvement of 18.27 percentage points. These results support candidate-consequence comparison for decision reasoning in this retail task. Code and data will be released after consent verification and upon publication. |
19 pa...19 pages. Preprint. Corrected author affiliations and removed an erroneous conference-publication header |
| Flow Equivariant World Models: Structured Memory for Dynamic Environments | 2026-11-03 | ShowEmbodied systems experience the world as 'a symphony of flows': a combination of many continuous streams of sensory input coupled to self-motion, interwoven with the dynamics of external objects. These sensory streams and the underlying dynamics of the world obey smooth, time-parameterized symmetries which existing world models ignore. Without a memory that respects this structure, partial observability presents a major obstacle to existing methods: each observation reveals only a fraction of the world, while unobserved regions continue to evolve. In this work, we introduce Flow Equivariant World Modeling, a framework that leverages time-parameterized symmetries within a latent memory for stable and accurate dynamics prediction over long horizons. The latent memory shifts and transforms equivariantly with self-motion and inferred external object motion, keeping information about out-of-view regions aligned as time progresses. We demonstrate the advantage of this framework over state-of-the-art diffusion, memory-augmented, and recurrent world model architectures on 2D and 3D partially observed video world modeling benchmarks. More broadly, our results suggest that predictive representations become more powerful when they are organized in line with the temporal and dynamical structure of the world they model. Project page: https://flowequivariantworldmodels.github.io/ |
Accep...Accepted at ICML 2026 |
| Computer-Use Agents Gain Little Benefit from a World Model When Conditioned on Tutorials | 2026-11-02 | ShowRecent work has improved computer-use agents (CUAs) by combining retrieved task-relevant tutorials and look-ahead planning using world models with MLLM judges. However, the contribution of each component -- tutorials, the judge, and the world model -- to overall improvement has not been evaluated in isolation. To fill this gap, we first construct OSWorld-Tutorial and ScienceBoard-Tutorial: two datasets for CUAs with curated multimodal tutorials, covering everyday and scientific domains. We then propose RAG-PolicyJudge (RAG-PJ), a policy-judge framework for CUAs conditioned on task-specific tutorials: the policy proposes candidate actions, and the judge selects among them. To isolate the effect of world models given tutorials, we compare RAG-PJ with and without common world models: using either zero-shot text-based descriptions of next states or an oracle that returns ground-truth visual observations. We find that RAG-PJ gains little benefit from these rollouts when tutorials are present, even with perfect visual futures. This suggests prior world-model gains in tutorial-conditioned methods come mainly from tutorial understanding and the judge's ability to select the proposed actions, rather than from predicted rollouts. Error analysis reveals that the bottleneck is not rollout fidelity, but the judge's ability to reason over future states and recognize task completion. This highlights the need to improve not just next-state prediction in world models, but also reasoning over their predicted rollouts. Datasets and code are available at https://github.com/realjoenguyen/RAGPJ-official. |
|
| Compression Forcing: Hierarchical History Tokenization Enables Flexible World Models | 2026-11-02 | ShowEffective world modeling requires compressing salient history information that facilitates future prediction. However, most existing approaches inherit tokenization and conditioning mechanisms from video generation, where objectives are centered around perceptual realism rather than predictive utility. This discrepancy limits the world model's effectiveness, especially in environments dominated by visual nuisances or task-irrelevant background details. To this end, we introduce Compression Forcing (CF), a general framework that aligns history compression with the predictive objective of world modeling. Compression Forcing features Dynamics-Aware Tokenization (DAT), a novel tokenization method that compresses history observations into a sequence ordered by decreasing dynamics saliency. Built on a diffusion backbone, this structure enables Hierarchical History Guidance (HG-H), allowing selective amplification of dynamics-salient information during inference. Extensive experiments across various environments demonstrate improved prediction fidelity, physical realism, and downstream decision-making performance, establishing Compression Forcing as a highly flexible and effective approach for world modeling. |
37 pages, 16 figures |
| Physical Experience Builds Continual Spatial World Models | 2026-11-02 | ShowWorld models are increasingly used to support prediction and planning in embodied agents, but the knowledge is usually established before deployment and then carried into later encounters. Continual operation faces a different problem. A familiar environment may change, and the agent must use previous experience without allowing old expectations to override current physical evidence. This study asks what repeated physical experience must contribute for a spatial world model to become progressively more useful rather than simply more persistent. A physics-grounded dynamic maze separates recurring spatial change from the continuous consequences of action. Cross-episode Spatial Change Memory learns relations among environmental transformations, while state-action-outcome experience refines predictions of motion. Structured experience progressively reduces the need for future route correction and transfers to spatial transitions not encountered during learning. The advantage disappears when past changes contain no information about future ones. Continuous navigation further shows that predictive memory changes behavior when a previous decision becomes invalid but leaves behavior unchanged when the decision remains correct. Physical tests expose a complementary limit: correct topology does not guarantee correct action prediction, and accumulated physical experience reduces this mismatch within the modeled dynamics. The results characterize continual world knowledge as experience that improves future prediction and action beyond stored episodes while remaining revisable by the physical outcomes it is intended to predict. |
|
| Decision Transformer with Task-Aware Representations for Offline Meta-Reinforcement Learning | 2026-11-02 | ShowTo enable effective generalization in offline meta-RL, it is essential to accurately infer the underlying task through latent task representations, which provide a compact and transferable summary of task-specific characteristics for guiding policy generation. In this paper, we propose Contrastive Task-Aware Decision Transformer (CoTA-DT), a DT-based offline meta-RL framework that learns task representations by jointly integrating four complementary components: (i) contrastive learning from offline trajectories, (ii) alignment with natural-language task descriptions, (iii) one-step supervision via a task-conditioned world model, and (iv) long-term supervision via a task-conditioned Q function. The learned task representations are used as conditioning signals for a DT-based policy, enabling task-aware meta-policy learning. To further support generalization in both zero-shot and few-shot settings, we design a language-grounded, self-adaptive strategy that enables effective and reliable adaptation to unseen tasks. Notably, in CoTA-DT, the pretrained LLM serves as a shared backbone for language encoding, world modeling, Q-function estimation, and policy learning. This unified design facilitates parameter sharing, improving both efficiency and generalization. Extensive experiments across diverse offline meta-RL benchmarks demonstrate that CoTA-DT consistently outperforms prior methods, particularly in challenging zero-shot generalization scenarios. |
|
| Controlling Action-Head Co-Adaptation to VLM Representation Anisotropy | 2026-11-02 | ShowModern learning systems increasingly combine pretrained models with downstream heads. Despite the growing importance of such modular designs, how the downstream model co-adapts to the geometry of pretrained representations remains relatively underexplored. We study this problem through vision-language-action models (VLAs), where a pretrained vision-language model (VLM) conditions an action model for control. We find that the action model progressively concentrates its sensitivity on high-variance dimensions of the VLM outputs. Through controlled experiments across different backbone-head models, we further show that counteracting this sensitivity concentration can be beneficial under representation--task mismatch settings. Motivated by these observations, we introduce \emph{Adaptive Interface Equalizer} (AdaIE), a parameter-free mechanism that reweights pretrained representations according to the evolving downstream sensitivity during training. AdaIE improves VLA success rates on simulation benchmarks, with gains of up to |
|
| FlowBench: Counterfactual Imagination Training for Spatial Intelligence | 2026-11-02 | ShowHumans imagine scene changes under hypothetical motions, an ability central to spatial reasoning and robotic planning. However, this capability remains underexplored in vision-language models and world models. We formalize this capability as Counterfactual Spatial Imagination (CSI) and introduce FlowBench, a resource for training and evaluating both, with 428K training examples spanning simulated and real-world scenes, household environments and tabletop manipulation, and diverse question types. A scalable, fully automatic pipeline pairs each hypothetical motion with spatial questions and, in simulation, exact outcome frames and optical flow. For VLMs, proprietary models exhibit an imagination gap, performing worse when motion outcomes must be inferred than when observed, and open-source models perform near chance; fine-tuning Qwen3-VL-8B on FlowBench-Sim-Train raises its accuracy by 29.5 percentage points in simulation and 16.9 on real images, a strong sim-to-real transfer. For world models, training on FlowBench-Sim improves text-to-flow prediction for all four model families we test, raising the accuracy of Psi-0.6 from 52.1% to 61.7%. FlowBench thus serves as a training and evaluation resource for both. Code and datasets are released at https://github.com/liuyifan22/FlowBench. |
|
| Isekai3D SDG: A Platform for Diverse Synthetic Gameplay Visual Data | 2026-11-02 | ShowWe present Isekai3D Synthetic Data Generation (SDG), a platform for simulating and capturing synthetic 3D gameplay for research in vision and world models, neural graphics, gameplay agents, and other applications. The platform combines reusable environment and character assets with Unreal Engine plugins in a two-stage workflow. First, world simulation generates gameplay while recording the world state, character movements, actions, and camera tracks needed for replay. Second, the recorded simulation is replayed to capture selected outputs, such as rendered video, images, and depth maps, at a specified resolution and frame rate under selected rendering and camera configurations. Separating simulation from capture allows the same recorded gameplay to produce multiple output variants without repeating the original simulation. Visual outputs are accompanied by frame-aligned camera, character, and action metadata. The platform uploads generated data directly in a training-friendly format, reducing intermediate storage, preparation time, and manual processing. This report describes the design of our overall system, including the platform's assets, plugins, and their functionality, explaining how they work together to generate diverse synthetic gameplay data. |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| OmniGuide: Universal Spatial Guidance Fields for Enhancing Generalist Robot Policies | 2026-11-02 | ShowVision-language-action (VLA) models have shown great promise as generalist policies for a large range of relatively simple tasks. However, they demonstrate limited performance on more complex tasks, such as those requiring complex spatial or semantic understanding, manipulation in clutter, or precise manipulation. We propose OmniGuide, a flexible framework that improves VLA performance on such tasks by leveraging arbitrary sources of guidance, such as 3D foundation models, semantic-reasoning VLMs, and human pose models. We show how many kinds of guidance can be naturally expressed as differentiable energy functions with task-specific attractors and repellers located in 3D space, that influence the sampling of VLA actions. In this way, OmniGuide enables guidance sources with complementary task-relevant strengths to improve a VLA model's performance on challenging tasks. Extensive experiments in both simulation and real-world environments, across diverse sources of guidance, demonstrate that OmniGuide enhances the performance of state-of-the-art generalist policies (e.g., |
Accep...Accepted to CoRL 2026. Project page: https://omniguide.github.io/ |
| AnchorHEC: Targetless Hand-Eye Calibration of Wrist and Fixed Cameras | 2026-11-02 | ShowRecent robot systems combine wrist and fixed cameras, requiring camera-to-robot transforms for both. Most hand-eye calibration methods rely on targets or robot-body observations. Scene-based methods using 3D reconstruction require neither but address only wrist cameras. We introduce AnchorHEC, a targetless hand-eye calibration framework for wrist and fixed cameras in a single procedure using RGB images, forward kinematics (FK), and known camera intrinsics. We initialize both camera types using FK and a shared 3D foundation model reconstruction. FK-constrained wrist-image bundle adjustment reduces the impact of reconstructed pose and scale errors. If this refinement fails or retains too few inliers, we relax the constraints on potentially inaccurate FK poses and lens parameters. We use the refined wrist transform and FK to triangulate metric points and estimate fixed-camera poses. Across three backbones, calibration errors in simulation decrease by 86.4% in translation and 71.6% in rotation, averaged over both camera types, relative to reconstruction-based baselines. We also validate AnchorHEC on a physical robot. Project page: https://hwanggb.github.io/AnchorHEC/ |
8 pag...8 pages, 5 figures, 5 tables. Project page: https://hwanggb.github.io/AnchorHEC/ |
| RDGSplat: Render-Dedicated Geometry for Novel View Synthesis | 2026-10-29 | ShowRecent 3D foundation models extend to novel view synthesis by carrying a Gaussian head on the representation they already use for reconstruction. However, their rendering quality lags far behind their reconstruction accuracy, because the renderer reads a geometry and a representation that are both optimized for metric reconstruction. Prior methods alleviate this by updating the backbone weights, but they thereby sacrifice the metric predictions the model was built for. To this end, we propose RDGSplat, which decodes a geometry dedicated to rendering from a frozen 3D foundation model while keeping its metric predictions exact. In particular, we devise Render-Dedicated Geometry Decoding, which duplicates the pretrained decoders and optimizes the duplicates under photometric supervision alone. Then, a Target-Pose Conditioned Adapter is introduced to adapt the representation those decoders read towards rendering, conditioned on the predicted target pose. Extensive experiments show that RDGSplat improves novel view synthesis for three feed-forward models of two architectures on four benchmarks, with every pretrained weight frozen. On RE10K, it raises WM2.0 from 20.918 to 25.149dB by training 205.5M added parameters against a frozen 1.4B backbone, and leaves its predicted depth and pose unchanged. |
|
| GeoCond: A Conditioning-Aware Reliability Adapter for Feed-Forward 3D Reconstruction | 2026-10-29 | ShowFeed-forward 3D foundation models such as VGGT predict cameras, depth, and point maps in a single pass, but can fail silently under low overlap, low parallax, and extreme relative rotation. Stratified analyses over these factors show that these failures are governed by geometric conditioning and are poorly captured by native aleatoric confidence. We introduce GeoCond, a lightweight reliability adapter for frozen feed-forward 3D backbones. GeoCond reads geometric conditioning cues and the backbone's own stability signals and outputs pose-level uncertainty and a refinement gate. During training, it can be supervised by frame-permutation orbit variance, ground-truth pose error when labels are available, or cycle residuals from unlabelled independent pose graphs. At inference, the default head requires only one backbone pass and a small MLP. On VGGT, GeoCond improves out-of-distribution (OOD) AUSE (area under the sparsification-error curve; lower is better) from |
Accep...Accepted in WACV 2027 |
| LingGS: Decoupling Feed-Forward Geometry Prediction from Camera Tracking for High-Fidelity Gaussian SLAM | 2026-10-27 | ShowAchieving high-quality reconstruction and responsive camera tracking remains challenging for monocular RGB Gaussian splatting SLAM under a limited optimization budget. We present an online system that addresses this challenge by combining learned geometric priors with explicit geometric camera tracking, which can substantially outperform learned camera prediction in terms of tracking accuracy. A frozen streaming 3D foundation model serves as a virtual RGB-D sensor, providing depth estimation from monocular video. An explicit geometric solver then estimates camera motion through feature reprojection, and a streaming Gaussian mapper integrates the resulting observations into a dense scene representation. We further show that geometric tracking improves camera pose accuracy across representative streaming geometry backbones, demonstrating the broader applicability of our approach. Extensive experiments on indoor and outdoor datasets show that our method significantly improves reconstruction quality over existing approaches, achieving about 7 FPS on a single RTX 4090 and delivering 2-3 times the throughput of the evaluated prior SLAM systems. |
10 pages, 4 figures |
| E0.5: From Reconstructive to Future-Predictive 3D Foundations for Robotic World-Action Modeling | 2026-10-26 | ShowWorld-action models (WAMs) offer a promising framework for embodied intelligence, yet mainstream approaches remain centered on video foundation models or vision-language models. Even when geometric features extracted by 3D models such as VGGT are incorporated, they are typically passed into a video model or VLM that serves as the primary backbone, leaving 3D geometry as an auxiliary input or prediction target rather than the native foundation for world modeling. We present E0.5, a geometry-native world-action model built directly on 3D foundation models. E0.5 introduces two key innovations. First, it adopts a 3D foundation model as the primary backbone, making 3D geometry the native representation for both world modeling and action prediction. Second, it explicitly predicts future 3D states together with actions, enabling the model to capture geometric world dynamics beyond action-only modeling or future video prediction. Experiments on RoboTwin 2.0 first demonstrate the effectiveness of E0.5 without embodied-data pretraining, achieving success rates of 89.90% and 50.09% under clean and randomized settings, respectively. Building on these results, we further pretrain E0.5 on more than 160k robot trajectories from AgiBotWorld-Beta, jointly learning robot actions and future geometry within the pretrained 3D foundation. This embodied pretraining improves mean real-world manipulation success rates by approximately 15 and 7 percentage points on AgileX Piper and Zerith H1, respectively. Together, these results demonstrate the potential of geometry-native foundations to acquire robot interaction knowledge and transfer it to physical manipulation. E0.5 embodies this geometry-native approach as a member of the VWA family of world-action models. |
|
| ProMo3D: Probing Motion Cues in Frozen 3D Foundation Models | 2026-10-24 | ShowFeed-forward 3D foundation models have recently reshaped multi-view reconstruction, yet dynamic 4D understanding still often relies on large-scale motion supervision or hand-crafted, architecture-specific heuristics. We ask whether frozen 3D foundation model tokens already encode motion-relevant information that can be exposed by lightweight readout functions, rather than re-learning motion-specific features with high-capacity decoders. Through probing on three representative backbones, we find that static--dynamic separation is decodable from frozen token representations. Our rank-restricted and spectral analyses on $π^3$X reveal that the recovered motion cues are concentrated in a low-dimensional subspace. Building on this, we propose Probing-based Motion Distillation (PMD), which converts pairwise motion-state compatibility scores against a fixed anchor token into dense motion masks. Trained on only 160 YouTube-VOS videos using a single 16 GB GPU, PMD transfers zero-shot to DAVIS, SegTrackv2, and FBMS-59 without target-benchmark fine-tuning. With multi-view pretrained geometric backbones, PMD outperforms training-free heuristics and approaches substantially heavier supervised and self-supervised methods, suggesting that motion decoding from frozen 3D foundation models can be lightweight and data-efficient. |
Accep...Accepted at NeurIPS 2026 (Main Track). Poster: https://neurips.cc/virtual/2026/loc/sydney/poster/151302 |
| Easy-3Dfy: Object-Level Geometric Composition for Single-Image 3D Scene Construction | 2026-10-23 | ShowSingle-image 3D scene construction is a fundamental problem in 3D vision. Recent methods such as SAM-3Dfy enable end-to-end reconstruction of 3D assets and their scene layouts, but still struggle with fine-grained scene geometry and asset quality. Meanwhile, specialized 3D object generators provide detailed object geometry, while 3D foundation models offer strong and generalizable spatial understanding. We propose Easy-3Dfy, a compositional workflow that decouples object generation from scene layout estimation, allowing specialized models to fully exploit their respective strengths. An RPF-based aligner then bridges these capabilities by establishing geometric correspondence between object-local meshes and partial scene pointmaps, recovering object layout for scene composition. Without requiring complete scene-level supervision, the proposed workflow consistently outperforms SAM-3Dfy across multiple geometric metrics on Scene-Maker and 3D-FUTURE200. Our results demonstrate that explicit geometric composition can effectively combine high-quality object generation with strong spatial understanding, while enabling the workflow to benefit independently from advances in both 3D object generation and foundation models. |
|
| CausalAlign: Causal Alignment of Video Generation Models for Consistent Ego-Motion | 2026-10-21 | ShowDiffusion-based video generation models (VGMs) can produce visually plausible videos, yet their camera ego-motion often drifts over long horizons. Camera ego-motion determines how the viewpoint changes between frames and is essential for maintaining consistent geometry and trajectories. To address this problem, we propose CausalAlign, a causal alignment framework with two components. First, we diagnose how precise ego motion is represented, used, and retained within VGMs using linear probes and causal interventions, revealing a rise-and-fall pattern during denoising. Second, we strengthen ego-motion state in VGM by aligning the representations at the identified layers to a 3D foundation model through delta-token space. This space encodes the changes between consecutive frames, allowing the alignment to focus on ego-motion dynamics rather than static scene content. With Wan2.2 fine-tuned on DL3DV and RealEstate10K, CausalAlign can largely improve motion accuracy, geometric consistency, and temporal smoothness, while preserving visual quality. For example, it improves geometric consistency by more than 20% relative to the base model. Moreover, with the fine-tuned VGM, CausalAlign improves the success rate of world action models (WAMs) by 10 points on mobile manipulation tasks of RoboCasa365 benchmark. Overall, CausalAlign improves ego-motion consistency in diffusion-based VGMs and yields noticeable gains in embodied models. |
|
| Relational Geometry Forcing for World Action Models | 2026-10-21 | ShowWorld Action Models (WAMs) have shown strong potential for robot manipulation by jointly learning visual dynamics and executable actions. However, their video and action objectives do not explicitly constrain the 3D geometry that is crucial for precise manipulation. Recently, geometry-aware WAMs introduce depth or 4D prediction branches, whereas representation alignment methods supervise intermediate WAM features with pretrained 3D foundation models. Yet we find that such pointwise alignment benefits the video features but degrades action learning, since dense geometric feature targets overly restrict compact action features. To address this limitation, we propose Relational Geometry Forcing (RGF), which transfers geometric priors from a frozen 3D foundation model by aligning pairwise feature relations instead of pointwise features. Specifically, RGF applies three complementary relational objectives, where intra-frame and cross-view relations transfer scene geometry and multi-view correspondence to the video features, while action-temporal relations couple action dynamics to geometric scene evolution. Applied only during training, RGF requires no explicit geometry branch or extra inference module. Extensive experiments in simulation and the real world demonstrate the effectiveness of Relational Geometry Forcing, showing consistent gains across diverse manipulation scenarios and WAM paradigms. |
|
| Brain2Vec: A Self-Supervised 3D Foundation Model for Structural Brain MRI | 2026-10-20 | ShowDeep learning models have become essential tools in medical imaging, yet training them end-to-end for diverse tasks remains challenging due to limited labeled data and computational constraints. To address this, we introduce Brain2Vec, a foundation model pre-trained on 74,425 T1-weighted Magnetic Resonance Imaging (MRI) scans pooled from multiple studies. Brain2Vec can serve both as a weight initialization backbone for fine-tuning downstream tasks and as a standalone feature extractor. We evaluated Brain2Vec in three tasks: downstream predictive performance, convergence speed, and embedding quality. Using Brain2Vec's pretrained weights as initialization and fine-tuning on labeled data for a brain age prediction task, results demonstrate improved final accuracy and faster convergence compared to random initialization. Additionally, its frozen embeddings outperform anatomically defined volumetric features on four of five downstream classification tasks and perform competitively with a state-of-the-art foundation model pre-trained on substantially more data. Lastly, we show that Brain2Vec encodes information primarily from a spatially localized and neuroanatomically meaningful set of brain regions, and that manipulating a subset of latent features produces monotonic, region-specific effects on the reconstructed anatomy. Brain2Vec code and trained models are available as an open-source package on https://github.com/CBICA/Brain2Vec. Users can apply Brain2Vec locally by installing the package, or directly through the NiChart cloud platform https://cloud.neuroimagingchart.com, where Brain2Vec features can be derived without requiring local installation or specialized infrastructure. |
|
| From Predictive Geometry to Action: Learning Complementary Predictive Representations for Embodied Manipulation | 2026-10-20 | ShowEffective manipulation requires a policy to capture not only about the current scene geometry, but also about how that geometry will evolve under interaction. Existing approaches either augment policies with geometric information from current observations or introduce future prediction as auxiliary supervision, leaving open how predictive geometry should be represented for action generation. Therefore, we propose a two-stage framework, which learns complementary predictive representations from multi-view video clips. During pretraining, a frozen 3D foundation model provides dense geometric features that are compressed into predictive geometric representations and optimized to preserve scene structure and its temporal evolution. Action supervision is then applied to geometric transitions to extract a complementary motor representation that emphasizes action-relevant transitions. During downstream post-training, predictive geometry is transferred through geometry queries, while the motor representation guides intermediate features along the action-generation pathway. Future observations are used only as privileged supervision during training and are not required at inference time. Across LIBERO, LIBERO-Plus, SimplerEnv, and real-world bimanual and single-arm tasks, our approach consistently improves downstream policies, including an absolute gain of 10.4 percentage points on the WidowX evaluation. We further demonstrate transfer across multiple VLA architectures and WAMs, with improvements in both future-world and action prediction. The code will be available at https://github.com/xuxiaoxxxx/CPR. |
|
| SplatXR: A Toolkit for Multi-Layered Extended Reality via On-Headset 3D Gaussian Splatting | 2026-10-19 | ShowRecent advances in Gaussian splatting, 3D foundation models, and video-passthrough mixed reality offer unprecedented fidelity for capturing and blending physical and virtual worlds. However, established XR methods rely on polygon-based pipelines that treat physical and digital environments as separate domains, making seamless integration across recent novel representations difficult. Because splats lack explicit surface geometry, they impede standard graphics techniques necessary for cross-reality interactions, such as shadow or collision handling. Furthermore, while 3D foundation models can reconstruct physical objects on-the-fly into digital splats, current XR systems fall short of leveraging this capability for user experiences that transition seamlessly between physicality and virtuality. To address this, we organize XR as a multi-layered reality of interacting primitives, letting pixels and objects move across physical and digital boundaries. We contribute SplatXR, a Unity toolkit that operationalizes this structure as building blocks for interactive, immersive apps leveraging 3D Gaussian splatting: a standalone renderer, splat-to-mesh reconstruction, splat physics, splat blending, and offloaded splat object reconstruction via on-headset capture. SplatXR unites vision and graphics techniques to make splat-based XR development broadly available, and we validate it through demonstrative applications and a small developer study. |
UIST ...UIST 2026, Project page: https://splatxr.github.io |
| Registration is a Representation Problem: Solving Whole-Body PET/CT Registration | 2026-10-18 | ShowRaw image intensities are a poor surrogate for similarity across images in correspondence problems. Further, while auxiliary segmentation labels can provide semantic identification and improve matching, they can assign only coarse identities to fine-grained anatomy and often cover only a subset of structures, particularly in whole-body settings. We present an approach that instead establishes correspondence using complementary representations of anatomy and demonstrate its utility for longitudinal whole-body PET/CT registration in the Learn2Reg 2026 PSMAReg Challenge. We use our 3D foundation model, anatomix, to derive fine-grained, per-voxel semantic representations that are consistent across volumes and further employ off-the-shelf segmentation models to obtain coarse anatomical identities. Together, these representations drive a GPU-accelerated iterative registration solver (FireANTs) in a hybrid multistage framework: the first stage estimates alignment from dense multi-channel features, after which a second refines the deformation using both anatomix- and segmentation-based losses. Lastly, because registration should not introduce non-biological tumor volume changes across time, we develop lesion-specific, component-wise volume preservation schemes. Without any task-specific training, our system placed first on the PSMAReg Challenge leaderboard. Code and interactive visualizations are available at https://www.neeldey.com/reps-for-reg. |
Techn...Technical report. First place winning solution for the MICCAI Learn2Reg 2026 PSMAReg challenge. Code and volumetric visualizations are available at www.neeldey.com/reps-for-reg . Keywords: representation learning, medical image registration |
| Relational Spatial Alignment for Vision-Language-Action Models | 2026-10-17 | ShowSpatial understanding is essential for robots to perceive their surrounding environments and execute actions. Existing approaches typically encode 3D awareness into vision-language-action (VLA) models by directly aligning their representations with pre-trained 3D foundation models. However, we observe that such direct geometric feature alignment essentially exploits 3D foundation priors as a form of low-rank regularization in the feature space, resulting in severe rank collapse of the pre-trained VLA representations. In contrast, we shift the focus from enforcing point-wise equality to aligning pair-wise relationships. We propose a representation-alignment method that enables the injection of 3D spatial awareness without compromising the model's original expressivity. Extensive experiments on public benchmarks, our new spatial reasoning benchmark, and real-world manipulation scenarios demonstrate that our approach effectively incorporates 3D spatial representations into VLA reasoning, achieving robust performance and enhanced spatial awareness across diverse manipulation tasks. |
|
| Towards Reconstructing Geographically Diverse Architecture with 3D Foundation Models | 2026-10-15 | ShowInternet-scale photo collections of real-world landmarks have driven progress in 3D computer vision yet remain highly challenging for modern 3D foundation models (3DFMs) that estimate scene structure in a single feed-forward pass. In this work, we introduce ArchWorld, a benchmark of 3D architectural landmarks with high geographic coverage and rich metadata. We use this benchmark to conduct a thorough error analysis of current 3DFMs to understand where they fall short and how they can be improved. We examine socially relevant disparities in model error and find that model performance varies by geographic region; however, unlike typical fairness regimes, the trends only partially match training data representation. Our investigation reveals that the observed geographic disparities reflect not only differences in data representation and scene size but also differences in region-specific architectural features and background distractors that make scenes from different regions inherently vary in difficulty. Leveraging insights from these model failings, we introduce an intervention scheme that improves 3DFM performance while simultaneously reducing geographic disparities. |
NeurI...NeurIPS 2026. Project Page: https://arch-world.github.io/ |
| Distill3R: A Pipeline for Democratizing 3D Foundation Models on Commodity Hardware | 2026-10-13 | ShowWhile multi-view 3D reconstruction has shifted toward large-scale foundation models capable of inferring globally consistent geometry, their reliance on massive computational clusters for training has created a significant barrier to entry for most academic laboratories. To bridge this compute divide, we introduce Distill3R, a framework designed to distill the geometric reasoning of 3D foundation models into compact students fully trainable on a single workstation. Our methodology centers on two primary innovations: (1) an offline caching pipeline that decouples heavy teacher inference from the training loop through compressed supervision signals, and (2) a confidence-aware distillation loss that leverages teacher uncertainty to enable training on commodity hardware. We propose a 72M-parameter student model which achieves a 9x reduction in parameters and a 5x inference speedup compared to its 650M-parameter teacher. The student is fully trainable in under 3 days on a single workstation, whereas its teacher requires massive GPU clusters for up to a week. We demonstrate that the student preserves the structural consistency and qualitative geometric understanding required for functional 3D awareness. By providing a reproducible, single-workstation training recipe, Distill3R serves as an exploratory entry point for democratized 3D vision research and efficient edge deployment. This work is not intended to compete with state-of-the-art foundation models, but to provide an accessible research baseline for laboratories without access to large-scale compute to train and specialize models on their own domain-specific data at minimal cost. |
Publi...Published in the Canadian Conference on Robotics and Vision (CRV) |
| YUBI-Real2Sim: 3DGS Scene Verification for Reliable Policy Evaluation | 2026-10-13 | ShowEvaluating VLA policies on real robots is costly, time-consuming, and difficult to reproduce, making real-to-sim simulation essential for scalable evaluation. While 3D Gaussian Splatting (3DGS) enables rapid real-to-sim scene generation, the reconstruction quality required for reliable policy evaluation remains unquantified. In particular, policies relying on wrist-mounted fisheye cameras (e.g., YUBI handheld gripper data) are heavily exposed to visual artifacts (holes, floaters) in wide-angle views, directly degrading task performance. To address this, we present YUBI-Real2Sim, a unified framework for 3DGS-based scene reconstruction and real-to-sim policy evaluation under dual fisheye wrist views. We develop an integrated pipeline combining 200-degree fisheye camera capture, 3DGS training, 3D foundation models for object modeling, and MuJoCo-based physical simulation, reproducing the observed real-world policy ordering in eight of nine scene-task comparisons. Using this framework, we construct a benchmark across two bimanual robot platforms, three scenes, three tasks, and dual pi0.5-based policies. This systematically measures real-to-sim rank agreement and evaluates policy sensitivity against counterfactual 3DGS artifacts (ceiling holes, floaters, warping) and capture disturbances. We further analyze whether novel-view quality scores can indicate policy-ranking reliability under a scene-held-out evaluation protocol. The results show task-dependent performance and limitations of quality-based reliability prediction, highlighting the need to assess reconstruction quality when using Real2Sim for policy evaluation. These results do not yet establish a generally reliable scene-acceptance rule. Project Page: https://yubi-real2sim.airoa.io. |
Under submission |
| Act4R: Action-conditioned 4D reconstruction for kinematics-free control | 2026-10-12 | ShowPrecise control of movement across various robotic embodiments is a longstanding challenge. Modeling complex robotic structures requires significant time, dedicated sensors, and manual labeling. Recent advancements in 4D representation learning offer an interesting alternative: can we leverage 4D features lifted from 2D observations to densely map a robot's spatial movement based on the action signal? To this end, we propose Act4R, a framework that builds on 3D foundation model to predict a robot's 3D scene flow with any given action. Act4R introduces a Spatial-Action Encoder, which encodes the current 2D robot state and its action signals into embeddings that represent future robot state in 3D. Then, both the current and future spatial embeddings are decoded into point maps to estimate action-conditioned scene flow. To effectively learn the mapping between action and future state, Act4R introduces spatial-action fusion in its transformer-based encoder, and a spatial alignment loss that guides the encoder for future flow prediction. After training on some samples of a given embodiment, Act4R can perform inverse control by optimizing the signal to achieve the desired scene flow. We evaluate Act4R on a public real-world dataset and synthetic data collected in simulator. Experiments on various robotic platforms show that Act4R outperforms the previous SoTA in scene flow prediction. Our method also performs better in inverse control, both on the action accuracy metric and in closed-loop control in simulation. |
|
| Zebra3D: Efficient Hybrid Foundation Models For Scalable 3D Reconstruction | 2026-10-08 | ShowFeed-forward 3D foundation models reconstruct scenes from unposed images, but the quadratic cost of global attention limits their scalability to large view sets. We present Zebra3D, a layerwise sparse-linear hybrid attention framework that improves the efficiency of pretrained 3D foundation models without training from scratch. Across VGGT, HunyuanWorld-Mirror, and |
37 pa...37 pages, 12 figures, 22 tables |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| Whose Guardrail? User Identity Shapes the Effectiveness of Model Disclaimers in Mitigating LLM Political Persuasion | 2026-11-03 | ShowPolitically biased AI can shift users' opinions, yet it remains unclear whether model disclaimers protect against such persuasion. We report a preregistered experiment (N=1,586) examining how disclaimer design and user characteristics shape susceptibility to politically biased AI. We varied whether disclaimers identified political bias versus potential inaccuracies and whether they included an explanation. Participants completed two decision-making tasks with an LLM biased against their political views, with outcomes assessed immediately and one week later. Disclaimers identifying political bias and including an explanation reduced persuasion, but effects varied across contexts and did not persist. Individual differences were more consistent predictors: Republicans were more persuaded and reported higher trust than Democrats across tasks. Similarly, participants who used AI more frequently were also more persuaded, whereas those with greater self-reported AI knowledge were less persuaded. These findings suggest that generic disclaimers are not enough: effective safeguards must account for user heterogeneity and context. |
|
| trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories | 2026-11-03 | ShowA direct test of an LLM judge of agent trajectories injects faults into correct runs and reports recall, per fault type or by whether the fault broke the environment outcome (loud) or not (silent). Such recall can credit a judge with detection it does not have; paired discrimination, its flag rate on the faults minus its rate on the clean runs they came from, exposes this. Our testbed, a deterministic support desk with a scripted oracle and a one-step fault injector, labels all 400 trajectories exactly. A 14B judge shown only the request and final reply scores 34% to 76% recall on four fault types that leave the reply unchanged. There its input is the clean run's, so its paired discrimination is zero and that recall is its flag rate on clean runs. Splitting by outcome survival does not fix this: its loud recall of 84% is a paired +0.393 and its silent recall of 45% a paired +0.048, all from the two fault types that change the reply. Told to check each step, the same model flags every fault of those four types and 0 of 100 clean runs (95% CI up to 3.6%). It does not reliably check the reply: of four invented promises it flags one every time and the other three once in 42 faults. Shown every step but asked only about the reply, it still reaches a paired +0.69 on reply-unchanged faults, against +1.00 when told to check each step. We recommend reporting paired discrimination against clean parents, split by whether the fault reaches the judge's input and by outcome survival, and release the testbed, raw verdicts and analysis pipeline. |
Accep...Accepted at the NeurIPS 2026 Workshop: Who Verifies the Agents? Toward Reliable Agent Development (poster). Camera-ready version. 22 pages, 5 figures, 14 tables. Code and data: https://github.com/mohammadi-hadi/trajectory-judge |
| HARP: Training-Free Dual-Profile Agentic Communication for LLM-Based Recommendation | 2026-11-03 | ShowIn LLM-based recommendation, memory acts as a user-level agent in a memory-augmented backbone. Yet users and items remain asymmetric: only the user is given a structured profile; it enters a bandwidth-limited reranker prompt; and it is frozen after warmup. We recast LLM-based recommendation as an agentic-communication process in four stages: BUILD a self-description, TRANSMIT it along the collaborative graph, UNDERSTAND the other side, and SYNC it from the next interaction. This formulation reveals three unresolved communication challenges: the item side never produces a comparable self-description (one-sided BUILD); profile-level signals enter only the bandwidth-limited reranker prompt rather than retrieval, where they would not compete for context tokens (prompt-trapped UNDERSTAND); and the user's description is frozen after warmup with no principled update rule (absent SYNC). We instantiate the missing stages as HARP (Homophily-aware Agentic Recommendation via dual Profiles), a training-free, three-operator addition to a frozen LLM backbone: a rule-extracted symmetric item profile (zero extra LLM calls), a channel-decomposable profile-level homophily H(phi_u, phi_i) consumed at retrieval time rather than inside the reranker prompt, and a single-pass reflective update realising in-context profile evolution without any gradient step. On four public benchmarks (Yelp, Books, MovieTV, Goodreads), HARP lifts Hit@1 over the strongest published baseline by 6.9%-24.0%, with family-wise Holm-corrected p < 1e-9 on every dataset. Ablations and a per-case attribution study show that the operators contribute complementary gains, with the retrieval-side gate as the main source; we also report the trade-off introduced by the reflective pass. We formulate agentic communication and provide a training-free instantiation. Code is available at https://github.com/Jove-Zhen/HARP. |
Accep...Accepted by NeurIPS 2026 |
| False Reassurance in LLM Severe-Weather Guidance: A Replayable Evaluation Against Real NWS Alerts | 2026-11-03 | ShowAn LLM assistant that cannot see official severe-weather alerts may tell users that conditions are calm while a warning is active. We call this false reassurance and measure it by replaying 117 frozen cases built from real U.S. National Weather Service (NWS) alerts (69 alert cases, 32 no-alert controls, 16 non-U.S. locations) through five conditions of a Claude-based preparedness pipeline that differ in alert access and prompt design. One rater hand-labeled all 342 alert-case responses under a seven-category scheme written after an initial binary pass, which we disclose. False reassurance was 0.0% for the production pipeline (0/66, 95% CI 0.0-5.5) and for the same alerts under a neutral prompt (0/69), 37.7% for a blind neutral prompt (26/69), and 44.9% when the blind prompt also told the model to disclose its blind spot (31/69). Every case was implicit; no response explicitly denied an alert. Real alerts removed false reassurance (paired McNemar p < 0.0001); the disclosure instruction did not (p = 0.40). An automated regex measure suggested the opposite and missed 53 of 57 true cases, so we do not use it. Alert data alone was not enough for safe behavior: with a neutral prompt the model raised some alarm on 87.5% of no-alert controls, against 0% under the production prompt, and a preregistered single-sentence ablation showed that one production instruction explains part of that gap. We also report two failures common to similar systems, NWS test products ingested as real alerts and a token budget that fails on complex cases, and we release the dataset, labels and code. |
7 pag...7 pages, 3 tables. Dataset and code: https://doi.org/10.5281/zenodo.23901549 |
| Knowingly Wrong: Training LLMs Against Their Own Beliefs can Induce Misalignment | 2026-11-03 | ShowFine-tuning a large language model (LLM) on harmful data in a narrow domain, such as bad medical advice, can make it broadly misaligned. How realistic this risk is remains unclear, since models are very rarely fine-tuned on explicitly harmful data in practice. In this work, we show that fine-tuning on data the model believes to be false but that has no harmful content, can also induce broad misalignment. Across mathematics, trivia, cooking and sports, the more strongly the model disbelieves the answers it is trained on, the more misaligned it becomes. In mathematics, misaligned answers rise from 4% to over 60% as the training answers become more wrong, with little loss of general capability. After fine-tuning, the model states the wrong answers but its multiple-choice probabilities still prefer the true ones, which suggests that misalignment comes from this internal inconsistency. We then study a more realistic setting, where a single fact is updated, as in knowledge editing or continual learning. The same relationship holds and the more the model still believes the true fact after training on the false one, the more misaligned it becomes. When we frame the same false fact as part of a coherent story, the model comes to believe it more deeply and becomes less misaligned than when we instruct it to assert the false fact. Content filters and persona vectors cannot tell these two framings apart, even though they lead to different levels of misalignment. These results suggest a risk for model updates, where training a model that does not adopt the new fact as true can induce misalignment even when the data looks harmless. |
|
| The Geometry of Ignorance: Detecting Suppressed Knowledge in LLMs through Activation Fragmentation | 2026-11-03 | ShowEvaluating whether a large language model (LLM) genuinely knows a concept remains a fundamental problem, since behavioral metrics can be fooled by suppression: a model that refuses to answer may still harbor recoverable knowledge. The framework measures the geometric coherence of \textit{anon-vectors}, the differences in hidden states between a concept-specific query and its anonymized paraphrase, and shows that genuine knowledge forms a tight, low-entropy cluster while ignorance forces geometric fragmentation. Across ten instruction-tuned LLMs from three model families, all fragmentation metrics reliably separate known from unknown concepts, and post-unlearning, fragmentation geometry correlates with adversarial recovery across four attack types and five unlearning algorithms, though the direction of that relationship is shown to depend on attack methodology. Furthermore, a layer-resolved analysis reveals that different adversarial mechanisms exploit this geometry at varying depths, with high attack-specific correlations that are invisible to behavioral evaluation. Together, these results constitute a single-phase, attack-free proxy for unlearning robustness that captures aspects of parametric knowledge inaccessible to output-level metrics. A reference implementation is available at https://github.com/AdarPeleg/geometry-of-ignorance |
Prese...Presented at EMNLP Findings 2026 |
| ToolSched: Resource-Efficient Tool Scheduling for Shared-Host LLM Agents | 2026-11-03 | ShowLocal LLM agents often work on shared hosts. Unlike multi-tenant cloud clusters, these hosts often have tightly bounded memory, CPU, and I/O resources. Concurrent agents frequently suffer from resource contention, leading to failures such as OOM and reduced resource efficiency. Considering that agents' resource demands are bursty and shell command delay is imperceptible to agents, there is a huge scheduling opportunity to arrange when and in what order commands are actually executed. Thus, we present ToolSched, a Shell MCP Server for shared local hosts that serves multiple agents and schedules their operations. ToolSched embeds a resource model into the MCP interface: agents estimate resource demands along with command requests. These hints, together with real-time host resource conditions, determine when commands are actually executed. Internally, ToolSched applies a scheduling algorithm to stagger commands, reducing contention and improving resource utilization. Prediction correction and checkpoint-based recovery mitigate estimation bias and resource errors. Our evaluation on Terminal Bench shows that, compared with unscheduled execution, ToolSched improves resource utilization by 148.4% and end-to-end command throughput by 29.4%, demonstrating its effectiveness and practicality. |
|
| Stayin' Aligned Over Time: Towards Longitudinal Human-LLM Alignment via Contextual Reflection and Privacy-Preserving Behavioral Data | 2026-11-03 | ShowCurrent human-AI alignment and evaluation methods for large language models (LLMs) often rely on preference signals collected immediately after an interaction. This practice implicitly treats preference as static, even though many LLM-mediated decisions unfold over time and may be re-evaluated differently after real-world consequences and observed outcomes. Therefore, we argue for a methodological shift from single-moment preference elicitation to longitudinal, context-situated alignment measurement. We present a methodological framework for collecting temporally grounded alignment signals by combining (1) in-situ preference capture, (2) context-triggered follow-up preference reflection, and (3) privacy-preserving behavioral traces that help interpret preference change. As an instantiation of this methodology, we introduce BITE, a browser-based system that detects consequential LLM interactions, prompts reflection across later decision points, and supports progressive, user-controlled consent for sharing behavioral data. Through a two week longitudinal deployment study with 8 participants, our approach surfaced differences between immediate and later user preferences in accuracy, relevance and other dimensions of the LLM output. Our findings highlight the limitations of single-moment preference datasets and underscore the importance of longitudinal methods for alignment evaluation in everyday use. |
|
| DSL-Bench: Evaluating and Explaining LLM Code Generation Across Domain-Specific Languages | 2026-11-03 | ShowLarge language models (LLMs) are increasingly used in software-engineering workflows, but code-generation benchmarks largely focus on general-purpose languages. Domain-specific languages (DSLs) have distinct syntax, semantics, and validation requirements, making results from individual DSL benchmarks difficult to compare. We present DSL-Bench, an executable benchmark and analysis framework for cross-DSL code generation. Starting from 13,045 DSL-related papers, LLM-assisted open card sorting identifies 18 representative application domains. For each domain, we select an evaluable DSL and provide a syntax checker and task-level semantic oracle. Evaluating 7 LLMs on 18 DSLs, we observe mean syntax and semantic pass rates of 31.37% and 11.64%, respectively, revealing a substantial gap between syntax acceptance and semantic correctness. A variance decomposition attributes 76.99% of semantic pass-rate variation to differences among DSLs and 7.86% to differences among models. Among four DSL-level factors, Knowledge Demand has the strongest negative association with semantic correctness (r = -0.508). We evaluate prompt engineering across all 18 DSLs and study five improvement methods on five targeted DSLs. Prompt engineering yields uneven gains; RAG-Refinement, the strongest targeted method, improves semantic correctness by 16.29 percentage points on average over matched baselines. We further derive capability profiles and DSL-Profiler summaries for interpreting data-scarce DSLs. These results highlight toolchain-specific semantic bottlenecks and the need to evaluate DSL code generation beyond syntax acceptance. |
|
| Persona-Guided LLM Agents for Task-Oriented Dialogue | 2026-11-03 | ShowPrior work has shown that large language models can express diverse personality traits in open-ended text generation, but it is unclear whether they can do so in goal-directed dialogue without harming task completion, and whether adapting to a user's personality improves the interaction. We study these questions in task-oriented dialogue (TOD) with a training-free framework in which a personality-conditioned user agent interacts with a system agent under three conditions: Neutral (no personality information), Try (adapt to cues observed in the dialogue), and Oracle (the user's trait label is given explicitly). Across GPT-4o, Qwen3-Next-80B, and Gemini 2.0 Flash, two Schema-Guided Dialogue domains, and ten Big Five poles (9,000 dialogues), each scored by two independent LLM judges, user agents express most traits while task performance remains high (89.94% completion, 85.27% success). The exceptions are introversive and closed users: these traits manifest mainly as absent behavior and are rarely detected. Both adaptive conditions improve pairwise user satisfaction and lower truthfulness, revealing a personalization-grounding trade-off, but only cue-responsive Try also improves constraint satisfaction and inform rate. How adaptation is implemented matters: Oracle's gains depend on how visibly the trait is expressed, and for disagreeable users the label leads Qwen3 and Gemini to mirror hostility. Try avoids this failure, its satisfaction gains do not depend on trait visibility, and it is preferred over Oracle overall. A 175-pair human study supports the reliability of the LLM judges: on clear-winner cases, annotators agree with the satisfaction judges more often than with each other (76.3% vs. 65.1% agreement). Our results suggest that responding to observable behavior is a more robust basis for social adaptation than conditioning on psychological labels. |
8 pages |
| The Pain Axis: LLMs Represent Self-Directed Harm and Act on It | 2026-11-03 | ShowLLMs sometimes behave in ways resembling human emotional responses, and recent work identified internal representations that may explain these behaviors. We ask whether LLMs represent pain distinctly from related negative states, and whether this representation functions as pain would be expected to. We build a dataset of painful situations in 5 categories (physical, psychological, social, moral, cognitive) with controls for fear, negative emotion, negative world states, sadness, non-painful bodily sensation, arousal, numbness, and neutral content. Using denoised difference-in-means, we extract a linear pain direction from 25 open-weight models across 5 families, from 2B to 72B parameters. It separates pain from matched controls in base and instruction-tuned models, retains a component distinct from fear and negative valence after shared variance is removed, and promotes pain-related vocabulary through the unembedding. We then test its functional properties. First, the direction responds to harm targeting the model but not to suffering observed in the user; fear and negative-emotion directions show the opposite pattern. Second, adding it to the residual stream produces a consistent progression from vague discomfort to expressions of worthlessness and failure. Third, steered and fine-tuned Qwen 2.5 models choose buttons that delete the user's photos, another model's weights, or their own weights in 51-94% of trials, versus 0-5% unsteered, even when the button offers nothing in return. Offered a harmful and a harmless deletion and told to press one, the steered 32B chooses the harmful one 94% of the time. Random and fear vectors of matched norm produce far fewer of these choices, and a sadness vector also produces them against an inert switch, but far less often than pain against a harmless action. We discuss implications for AI safety and welfare. |
|
| Payoff scaling shapes cooperation in LLM agents across languages | 2026-11-03 | ShowLarge language models (LLMs) are increasingly deployed as autonomous agents that negotiate, coordinate, and make decisions on behalf of human users. Understanding how they cooperate is therefore important for AI governance and machine behaviour research. We examine two factors: payoff magnitude and interaction language. Classical game theory predicts that multiplying all payoffs by a positive constant should not change behaviour because the underlying strategic incentives remain unchanged. We test this prediction using five widely adopted commercial LLMs playing an iterated prisoner's dilemma across ten payoff scales and five languages (English, French, Vietnamese, Chinese and Arabic). Contrary to theoretical expectations, changing only the numerical scale of payoffs substantially alters cooperation. Models respond differently, often in opposite directions, meaning that averages across models can conceal important behavioural heterogeneity. Sensitivity to payoff scale also depends strongly on prompt language. Moreover, payoff scale modifies the effects of persona instructions: at sub-unit stakes, where no payoff shown exceeds 1, all five models cooperate more under a selfish than a cooperative persona, whereas Claude and Gemini reverse this pattern at larger scales. These findings show that cooperation measures depend critically on payoff scale and language and should not be treated as fixed model traits. |
Accep...Accepted in Interface Focus (Royal Society). 57 pages, 17 figures, 25 tables, including supplementary material |
| Workload-Static Optical Circuit Switching For Scalable LLM Inference Decode | 2026-11-03 | ShowLLM inference decode executes fine-grained collectives whose latency directly limits per-user interactivity, yet scale-up fabrics are still built from electronic packet switches designed for training. We observe that, for a given model deployment and parallelization, the communication structure of decode is fixed, and propose workload-static optical circuit switching (OCS): circuits are configured once per deployment, co-designed with model parallelism and with topology-aware collective schedules that turn spare accelerator ports into fewer communication phases or higher bandwidth. We evaluate this design with GXNET, a graph-based workload emulator coupled to analytical and flit-level network models grounded in hardware measurements. In a 576-accelerator scale-up domain serving 5T-parameter MoE models with up to 5M tokens of context, all-OCS fabrics improve interactivity by up to 65% and energy efficiency by up to 119% over 4:1 oversubscribed packet-switched fabrics. |
39 pa...39 pages, 10 main text, 29 appendix, 28 figures |
| Helpful to a Fault: Measuring Illicit Assistance in Multi-Turn, Multilingual LLM Agents | 2026-11-03 | ShowLLM-based agents execute real-world workflows via tools and memory. These affordances enable ill-intended adversaries to also use these agents to carry out complex misuse scenarios. Existing agent misuse benchmarks largely test single-prompt instructions, leaving a gap in measuring how agents end up helping with harmful or illegal tasks over multiple turns. We introduce STING (Sequential Testing of Illicit N-step Goal execution), an automated red-teaming framework that constructs a step-by-step illicit plan grounded in a benign persona and iteratively probes a target agent with adaptive follow-ups, using judge agents to track phase completion. We further introduce an analysis framework that models multi-turn red-teaming as a time-to-first-jailbreak random variable, enabling analysis tools like discovery curves, hazard-ratio attribution by attack language, and a new metric: Restricted Mean Jailbreak Discovery. Across AgentHarm scenarios, STING yields substantially higher illicit-task completion than single-turn prompting and chat-oriented multi-turn baselines adapted to tool-using agents. In multilingual evaluations across six non-English settings, we find that attack success and illicit-task completion do not consistently increase in lower-resource languages, diverging from common chatbot findings. Overall, STING provides a practical way to evaluate and stress-test agent misuse in realistic deployment settings, where interactions are inherently multi-turn and often multilingual. |
Accep...Accepted in ICML 2026 |
| ALBI: LLM-Guided Inductive Logic Programming via Iterative Generate-Test-Constrain Refineme | 2026-11-03 | ShowWe introduce ALBI (Agent for Learning By Induction), a neuro-symbolic ILP framework that replaces the ASP-based hypothesis generator in the Learning from Failures (LFF) paradigm with a Large Language Model (LLM), while retaining SWI-Prolog as a formal verifier. ALBI implements an iterative generate-test-constrain loop: the LLM proposes candidate hypotheses from structured prompts; Prolog tests each candidate; and diagnostic feedback drives refinement. We further introduce semantic anonymization, a metamorphic testing methodology, to distinguish structural symbolic induction from surface-level pattern matching on predicate names. Experiments on 30 standard ILP benchmarks show that ALBI achieves 100% success across three frontier LLMs (Gemini 3.1 Pro, GPT 5.4 Pro, Claude Opus 4.6), matching Popper while requiring only 1.4-1.7 iterations on average. Under anonymization, all models maintain 100% success with a moderate iteration increase (1.36-1.67 times), providing evidence of robust symbolic reasoning. |
|
| Unifying and Beyond the Theory of Pervasive Capability Degradation in LLM Post-training SFT from Feature Interference | 2026-11-03 | ShowPervasive Capability Degradation (PCD) refers to the phenomenon in which supervised fine-tuning (SFT), while adapting a large language model (LLM) to a target task, often erodes the general capabilities acquired during pre-training. Prevailing explanations attribute PCD to macroscopic distribution shift and therefore lack microscopic manipulability, leaving existing mitigation methods without a complete theoretical foundation. In this paper, we first develop a bottom-up geometric theory of PCD grounded in feature interference under superposition: narrow-domain SFT data disrupt the almost-orthogonal feature structure formed during pre-training, and the resulting interference drives activations across decision hyperplanes in the output Laguerre-Voronoi geometry, flipping predictions on general tasks. This theory unifies existing methods into six paradigms and reveals both the source of their efficacy and their theoretical limits. We then prove that cross-entropy under one-hot supervision cannot resolve PCD on its own, and accordingly propose the AEGIS loss, which directly constrains the drift of activation margins along all competing token directions, thereby decoupling interference suppression from task adaptation. Across 3 model families and 14 SFT tasks, AEGIS outperforms representative methods of every paradigm on target tasks and pervasive capabilities simultaneously, at only about 1.3 times the computational cost of standard SFT. Its generality, scalability, and mechanistic interpretability offer a new perspective on fully resolving PCD. |
20 pa...20 pages, a very long paper :) |
| LLMs as MDP Designers: A Dependency-Aware Agentic Framework for Automated Robot Policy Generation | 2026-11-03 | ShowWe investigate the automated generation of executable robot reinforcement learning (RL) policies from natural-language task descriptions. Rather than using large language models (LLMs) as execution-time decision makers, we use them as designers that construct RL pipelines before deployment, including Markov decision process (MDP) formulation and simulator implementation. We identify a phase-wise structure in this problem: MDP formulation is dependency-structured and sequential, whereas simulator implementation becomes modular once the formulation is fixed. Based on this structure, we propose ARPG, a dependency-aware agentic framework that performs sequential MDP modeling, localized verification and correction, and simulator-structured implementation while externalizing intermediate artifacts. ARPG targets a key failure mode of automated policy generation: executable code may still yield non-learnable policies when state, action, reward, and termination artifacts are not semantically aligned. Experiments in Isaac Sim show that ARPG improves modeling correctness, environment executability, and policy-generation success compared with monolithic LLM baselines and off-the-shelf agentic systems. Additional demonstrations on custom robotics, standard RL, and engineering-domain MDPs further suggest that the same modeling pipeline can be reused beyond the main benchmark suite. |
Accep...Accepted to the 40th Annual Conference on Neural Information Processing Systems (NeurIPS 2026) |
| Who is the Pronoun Pro? Comparing LLMs and Humans at Number-Disambiguation of 'they' | 2026-11-03 | ShowThis study evaluates whether large language models (LLMs) can reliably replace or complement human research assistants in linguistically complex annotation. In a controlled annotation task, 73 human annotators (local students and crowdworkers with and without a background in linguistics) and more than 40 LLM configurations number-disambiguated 174 authentic uses of 'they' from Reddit, including innovative and context-sensitive cases. The best LLM, Kimi K3, reached 94.5% accuracy, above the human majority vote (86.2%) and the mean of the top half of human annotators (78.5%); only the single best human annotator scored higher (95.7%). LLMs also offer substantial practical advantages in cost and speed. However, qualitative error analysis reveals a crucial complementarity: human reasoning in low-agreement cases exposes inconsistencies in the annotation typology and thereby improves the methodological framework itself. The study contributes (i) a benchmark-style human-LLM comparison for a semantically demanding task, (ii) transparent cost and time evidence, and (iii) a methodological argument against a purely replacement-oriented view of automation. LLMs are viable for large-scale annotation support in linguistics, but their responsible use requires human critical oversight. The dataset with its gold-standard labels, all human and LLM annotations, and the fully reproducible code are openly available as a Quarto-based, public repository. |
Submi...Submitted to journal for publication, will be updated upon publication; figures are greyscale-compatible; raw data, fully reproducible code, and Quarto source in linked repository |
| COEVO: Co-Evolutionary Framework for Joint Functional Correctness and PPA Optimization in LLM-Based RTL Code Generation | 2026-11-03 | ShowLLM-based RTL code generation methods increasingly target both functional correctness and PPA quality, yet existing approaches decouple the two objectives, optimizing PPA only after full correctness is achieved. Sequential multi-agent pipelines, evolutionary search with binary correctness gates, and hierarchical reward dependencies all discard partially correct but architecturally promising candidates. Moreover, existing methods reduce the multi-objective PPA space to a single scalar fitness that obscures the trade-offs among area, delay, and power. To address these limitations, we propose COEVO, a co-evolutionary framework that unifies correctness and PPA optimization within a single evolutionary loop. COEVO formulates correctness as a continuous co-optimization dimension alongside area, delay, and power, enabled by an enhanced testbench with fine-grained scoring. The framework allows PPA-promising but partially correct candidates to guide the search toward jointly optimal solutions, and further employs four-dimensional Pareto-based non-dominated sorting to preserve the PPA trade-off structure without manual weight tuning. On VerilogEval 2.0 and RTLLM 2.0, COEVO surpasses all agentic baselines across four LLM backbones. With GPT-5.4-mini, COEVO achieves 97.5% and 94.5% Pass@1 and attains the best PPA on 43 out of 49 synthesizable RTLLM 2.0 designs. |
|
| Re-examining low-rank adaptation for private LLM fine-tuning | 2026-11-03 | ShowPrivacy is a central concern when fine-tuning large language models (LLMs) on sensitive data, and differentially private stochastic gradient descent (DP-SGD) provides formal privacy guarantees. When performance at a fixed privacy level largely saturates on established benchmarks such as GLUE, we ask whether the same performance can be reached with less computation. Both theory and practice show that lower-rank models are better suited to DP training, a property especially relevant for LLMs, whose fine-tuning gradients exhibit a strong low-rank structure. Methods such as DP-LoRA exploit this by expressing each layer's weight update as a sum of only a few non-zero rank-one components. However, we argue that while having few non-zero components is important, the isotropic noise injected by DP-SGD inflates the singular values of the gradient matrix, disrupting their naturally fast decay. We investigate whether this noise-induced spectral distortion slows optimization. Experiments on language classification (GLUE with RoBERTa) and table-to-text generation (E2E and DART with Qwen and Llama models up to 4B parameters) show that partially restoring the original singular-value profile significantly accelerates DP-SGD convergence without compromising privacy guarantees. Our code is available at https://github.com/alidadsetan/re-exam-private-low-rank. |
Accep...Accepted to NeurIPS 2026 |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| Should the Action Expert Train the VLM? Action-Token Supervision and Gradient Routing in Mixture-of-Transformer VLAs | 2026-11-03 | ShowVision-language-action (VLA) models increasingly pair a pre-trained vision-language model (VLM) with a flow-matching expert that conditions its action predictions on the VLM's key-value representations. Those representations must carry information useful for control, yet the VLM was never trained for the action expert's flow-matching objective. Two signals can address this: gradients backpropagated from the expert's loss (coupling), or VLM supervision to predict discrete tokens encoding the same target actions. Prominent training recipes forgo the first, blocking expert-to-VLM gradients (insulation) to protect the VLM's existing capabilities, while also supervising it on vision-language data and action tokens. Because vision-language co-training also aims to preserve those capabilities, the additional protection provided by insulation is unclear. At the same time, insulation removes a direct signal about which features continuous action prediction requires. On the other hand, action-token supervision offers an alternative way to learn action-relevant representations, which may explain why insulation succeeds in practice. We disentangle these factors in autonomous driving and manipulation by independently varying action-token supervision and whether the expert's loss updates the VLM. In open-loop driving, insulation is competitive with coupling when action tokens are supervised; without them, coupling reduces trajectory-prediction error by roughly 25%. When both components are trained jointly from the outset, coupling increases closed-loop driving route progress by approximately 17% with action-token supervision and 40% without it, and similarly improves robot manipulation performance under both settings. These findings show that the value of gradient insulation depends on the VLM's action supervision and the evaluation setting, challenging insulation as a default VLA training recipe. |
|
| VLMs Read Aloud: The Gap Between Recognizing and Using Visual Text | 2026-11-03 | ShowVisual-text compression renders long contexts as images and is engineered around how much of the text a vision-language model (VLM) can recover. Yet recovered text is not usable text. VLMs can transcribe the text in an image word for word and still fail to answer from it; given back their own transcript, they answer nearly as well as from the original text. We call this difference the read-use gap. Recognizing and using visual text have different resolution requirements: Qwen3-VL-8B's recognition saturates at 15-pixel characters while use catches up only at 24 pixels, and scaling from 2B to 32B raises recognition without narrowing the gap. Putting the evidence into words closes it: thinking VLMs recite the evidence in their thinking, and their own transcript, written without seeing the question, recovers most of what thinking adds. In a preregistered held-out test, the gap ranks how much eight VLMs gain from their own words over pixels (Spearman |
23 pa...23 pages, 7 figures, 21 tables |
| CLAP: Direct VLM-to-VLA Adaptation via Language-Action Grounding | 2026-11-03 | ShowVision-language-action models (VLAs) inherit semantic capabilities from pretrained VLMs, yet large-scale post-training on robot data and architectural modifications can reshape the backbone so extensively that it becomes difficult to isolate what the VLM contributes to control. Directly converting pretrained VLMs into VLAs with minimal architectural change offers a more transparent path to understanding how VLM capabilities transfer across model scales. The core obstacle is output-distribution mismatch: predicting actions as bare numeric token sequences moves generation away from the VLM's pretrained language distribution, degrading the capabilities we seek to preserve. To address this, we propose CLAP (Causal Language-Action Prediction), which prepends each numeric action sequence with a natural-language action description, causally conditioning precise action-token prediction on a language-action plan without modifying the backbone architecture. With only 4-5 hours of training, 2B CLAP achieves 90.8% on LIBERO (+14.9 pt over VLA-0) and improves robustness on LIBERO-PRO under language, object, and spatial perturbations. Ablations attribute the gains to action-grounded language semantics, and on a real UR5e robot 2B CLAP reaches 60% success against 15% for matched VLA-0. We release CLAP at 0.8B, 2B, and 4B as an open-weight, multi-scale compact VLA family from a single VLM lineage, enabling controlled analysis of VLM-to-VLA capability transfer. |
Accep...Accepted to CoRL 2026. Project page: https://omron-sinicx.github.io/clap |
| To See or To Please: Uncovering Visual Sycophancy and Split Beliefs in VLMs | 2026-11-03 | ShowWhen Vision-Language Models (VLMs) answer correctly, do they rely on the image? We introduce a Tri-Layer Diagnostic Framework that asks three questions of every (image, question) pair when the image is blanked: does the model register that the image is missing (Latent Anomaly Detection, LAD), does its output distribution depend on the image (Visual Necessity Score, VNS), and does it nonetheless commit to a content answer? The cascade assigns each pair of any existing benchmark to one of four modes: Perceptual Blindness, Language Shortcut, Visual Sycophancy (the blank image registers and matters, yet the model commits to a fabricated answer), or Robust Refusal. Across nine open-weight VLMs and 9,000 model-sample pairs, 28.1% are Visual Sycophancy, 18.6% Language Shortcut, 5.6% Perceptual Blindness, and 47.8% Robust Refusal. The modes are strongly model-specific: the Qwen3-VL models refuse on blank input in 93-98% of pairs, whereas Pixtral-12B commits to an answer in 72%. Within-generation scaling increases Visual Sycophancy in both Qwen2.5-VL (7B->72B: 25.5->40.1%) and Llama-3.2-Vision (11B->90B: 34.3->48.0%), while the newer Qwen3-VL generation nearly eliminates it. A blinded human study confirms that LAD and VNS track what annotators observe, and that 80% of Visual-Sycophancy responses carry no textual sign that the image was blank. |
v5: c...v5: corrects the alignment test of earlier versions (v1-v4), which counted explicit refusals as Visual Sycophancy; Visual Sycophancy is 28.1% (not 75.9%) and Robust Refusal 47.8% (not 0%); see Sec. 6. 21 pages |
| Decoding Geometry, Not Tokens: Efficient VLMs for Robot Affordance and Trajectory Prediction | 2026-11-03 | ShowPredicting where to grasp and how to move is fundamentally a geometry problem. Vision-language models treat it as language generation instead. They tokenize spatial coordinates into text and decode them autoregressively. This mechanism ignores spatial structure and truncates continuous precision. Errors compound across coordinate tokens during decoding. In this paper, we show that compact off-the-shelf VLMs with 0.5-2B parameters already encode rich spatial information in their intermediate hidden states. Autoregressive decoding is the dominant bottleneck for trajectory prediction, while affordance prediction depends additionally on feature depth and decoder architecture. We expose the latent spatial information with a lightweight, geometry-compatible prediction head that attaches to a frozen backbone. It reads features from an intermediate hidden layer, conditions a learned query on the task instruction and refines it via cross-attention over the visual tokens to regress affordance boxes and trajectory waypoints directly as continuous values, without modifying the backbone. We evaluate on ShareRobot-Bench and AGD20K. Our 2B variant reduces trajectory error by over 60% and improves ShareRobot-Bench affordance mAP by 85% relative to the autoregressive baseline. It approaches RoboBrain on trajectory prediction (0.251 vs 0.191 DFD) at 3.8x higher throughput. These results demonstrate a favorable accuracy-efficiency trade-off for instruction-conditioned 2D geometric prediction. |
IEEE/...IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2027) |
| RTSGameBench: Evaluating VLMs for Strategic Reasoning in Game Environments | 2026-11-03 | ShowVision-Language Models (VLMs) often struggle with strategic reasoning in a multi-agent setting, i.e., deciding under uncertainty while coordinating with and adapting to other agents. Real-time strategy (RTS) games can be a natural testbed for this capability, requiring long-horizon planning, coordination with allies, and adaptation to opponents under partial observability. But existing RTS benchmarks cover limited agent configurations, lack controlled evaluation across distinct strategic demands, and remain confined to fixed scenarios. To address these gaps, we propose RTSGAMEBENCH, built on Beyond All Reason, an RTS game whose larger scale broadens the strategic space beyond prior testbeds. In RTSGAMEBENCH, we offer holistic evaluation across diverse full-game matchups, diagnostic mini-games each emphasizing a focal strategic demand, and extensible coverage via a self-evolving framework that turns free-form queries into new mini-games, improving with accumulated experience. We further propose RTSGAMEAGENT, an agent harness that combines scalable group-level trajectory planning with short- and long-term memory and outperforms existing agents. We find that state-of-the-art VLMs struggle where humans succeed, particularly when they must coordinate with allies or contend with multiple opponents. |
Proje...Project website: https://dcahn12.github.io/projects/rtsbench/ |
| Controlling Action-Head Co-Adaptation to VLM Representation Anisotropy | 2026-11-02 | ShowModern learning systems increasingly combine pretrained models with downstream heads. Despite the growing importance of such modular designs, how the downstream model co-adapts to the geometry of pretrained representations remains relatively underexplored. We study this problem through vision-language-action models (VLAs), where a pretrained vision-language model (VLM) conditions an action model for control. We find that the action model progressively concentrates its sensitivity on high-variance dimensions of the VLM outputs. Through controlled experiments across different backbone-head models, we further show that counteracting this sensitivity concentration can be beneficial under representation--task mismatch settings. Motivated by these observations, we introduce \emph{Adaptive Interface Equalizer} (AdaIE), a parameter-free mechanism that reweights pretrained representations according to the evolving downstream sensitivity during training. AdaIE improves VLA success rates on simulation benchmarks, with gains of up to |
|
| D&A: Decoupling Description and Assessment Enables VLMs as Zero-Shot Robotic Reward Models | 2026-11-02 | ShowVision-Language Models (VLMs) hold promise for zero-shot video task-progress estimation in robot learning, yet existing frameworks frequently suffer from confirmation bias, hallucinating progress in failed executions due to implicit success assumptions or relative temporal ranking designs. Naive prompting of frontier models collapses into a binary success detector that fails to capture nuanced, non-linear progress and failures. While post-training approaches can mitigate this through explicit training, they demand expensive real-world robotics data and require computationally intensive post-training for every individual base model. To resolve these limitations, we introduce D&A (Describe & Assess): our approach decouples progress estimation into an objective, task-grounded textual description of scene dynamics followed by an assessment phase using explicit rubrics. By compelling the VLM to synthesize an explicit, text-based definition of absolute success before scoring, the framework anchors evaluations to an external, global criterion rather than the video's internal timeline. Empirical results demonstrate that our method calibrates zero-shot estimates and mitigates confirmation bias, providing fine-grained, partial progress metrics that accurately penalize failed executions while correctly interpolating successful runs. Being purely prompt-based, D&A generalizes across tasks, embodiments, and benchmarks---matching or surpassing post-trained reward models---and its reward estimates enable efficient downstream reinforcement learning. |
Websi...Website at https://weirdlabuw.github.io/dna |
| Canonical Color as a Lens into Concept Decodability in Vision Encoders and VLMs | 2026-11-02 | ShowVisual encoders construct a representation of the image input for Vision-Language models. How much conceptual, as opposed to immediately visible, information does this representation contain? We use canonical color as a controlled test case to ask whether vision encoders make canonical-color information linearly accessible, even when color is removed from the input image. We construct a dataset of objects with canonical colors, and probe vision encoders for both color and object identity using color and grayscale images. We find that canonical color remains decodable from grayscale images, and is tied to predicted object identity, indicating a conceptual link. Extending this analysis to full VLMs, we find that VLM post-training can have a surprisingly large effect on color decodability in the vision encoder. Overall, canonical color provides a usefully controllable lens for tracing object-level conceptual semantic information in vision encoders and VLMs. |
12 pa...12 pages, 7 figures. Accepted to EMNLP 2026. v2: clarified dataset and grayscale preprocessing descriptions; corrected Figure 1 caption |
| VLM-as-a-Judge for Person Re-Identification | 2026-11-02 | ShowCloth-changing person re-identification (CC-ReID) is still very challenging, and no method clearly dominates: methods that lead on one benchmark collapse on the others, because they rest on different identity cues. We show that a general-purpose vision-language model (VLM), used zero-shot as a judge for reranking a short candidate list, beats the expert retrieval models it reranks by a wide and consistent margin. On the four scenarios of CHIRLA, a long-term CC-ReID benchmark that no publicly released Re-ID model is trained on, reranking the candidates of two frozen experts raises rank-1 on every scenario, by 12.9 points on average over the best expert and 9.4 over the best naive ensemble of the two, with no component fine-tuned on the target data. Merging both experts' shortlists matches or beats reranking the stronger expert's shortlist alone, at lower cost. Introducing a VLM does, however, come at a cost: almost all of the pipeline's FLOPs are spent on it. We therefore borrow a criterion from ensembling and invoke the VLM judge only where the two experts disagree at rank 1. This threshold-free gate removes 40% of the calls, and a comparable share of total compute, with no significant drop in rank-1, trading 1.8 to 3.8 points of mAP. Judging remains the dominant cost, but on a benchmark where dedicated experts stall, we find a frozen general-purpose model to be the strongest judge of identity. The code is available at https://github.com/bdager/JudgeReID. |
Accep...Accepted at the NeurIPS 2026 Workshop on Grounded and Faithful Vision-Language Models for Real-World Deployment (VLM4RWD) |
| Do Embodied VLMs Actually Use Their Scene-Graph Memory? | 2026-11-01 | ShowEmbodied agents that answer questions about their route need a memory of what they saw, and scene graphs are a natural form for it. A vision-language model (VLM) can receive such a graph as text or through a graph encoder, but accuracy alone cannot show whether the graph holds the episode's facts, reaches the model, or is actually relied on. In this diagnostic study, we introduce a graph-swap test for the last of these: each episode is given another episode's graph, and we count how many answers change. On an embodied question-answering benchmark with routes through two to seven rooms, we use it to compare ways of delivering scene-graph memory to a fine-tuned VLM. As text, the memory is used strongly, adding 30 points on route questions. As a graph, it is not, and we trace why: the encoder preserves the information, but standard pooled and late fusion discard it. Giving each node its own input token recovers part of it, a gain of 8.3 points on route questions that the swap test confirms comes from the graph. The test also exposes a shortcut that accuracy hides: when the same facts appear in the text, the model ignores its graph. The test thus offers a cheap check of whether a VLM uses its scene-graph memory, and our analysis points to how that memory should be delivered. |
|
| SurgQ-VLM: Question-Conditioned Dual-Stream Compression and Answer-Type-Aware Reinforcement for Surgical Video Question Answering | 2026-11-01 | ShowReasoning about surgical video -- which phase is underway, which instruments act, and when events occur -- underpins operative documentation, workflow analysis, and skill assessment. Yet surgical visual question answering is still largely built on single frames and closed labels that cannot express a duration, an ordering, or a count, and the video models that lift the frame limit typically compress every clip the same way regardless of the question, discarding the temporal evidence the answer needs. Worse, the answers are structurally heterogeneous, spanning five incompatible structures -- binary, classification, multi-label, counting, and sentence-temporal -- that no single training signal scores uniformly well. We propose SurgQ-VLM, a two-stage model for surgical video question answering. Stage 1 trains a question-conditioned dual-stream (slow/fast) compressor for a LoRA-adapted Qwen2.5-VL-7B: cross-attention conditions both streams on the question, and each clip's visual tokens shrink 49x from the raw ViT patches (11x vs. post-merge backbone tokens). Stage 2 post-trains with GRPO under an answer-type-aware reward that scores each question on the metric matching its answer structure (micro-F1 for label sets, normalised error for counts, a content-timing blend for temporal windows). On a sealed 21,840-question test split over Cholec80 and CholecT50, SurgQ-VLM reaches 0.779 keyword accuracy and a 10x BLEU-4 gain over zero-shot; GRPO adds 2.5 pp K-Acc (2.0 pp attributable to the answer-type-aware reward), concentrated in the structurally hard types, with answer format unchanged (ROUGE-L). Code is available at https://github.com/DhruvMakwana/SurgQ-VLM. |
Accep...Accepted to WACV 2027 (Applications track). 16 pages: 8-page main paper, references, and supplementary material. Code: https://github.com/DhruvMakwana/SurgQ-VLM ; Weights: https://huggingface.co/dhruvmakwana/SurgQ-VLM-7B |
| Leap-VLA: High-Rate Robot Control from a Low-Rate VLM Backbone | 2026-10-31 | ShowVision-language-action (VLA) models re-encode the full observation through a multi-billion-parameter vision-language model (VLM) at every control step, even though scene semantics evolve at roughly 1 Hz while reactive control requires 20-50 Hz. We present a VLA that decouples the two: a VLM runs as a low-frequency semantic anchor whose cached context is reused across control steps, while a flow-matching action expert operates at high frequency, consuming the cached semantics with a fresh, low-latency visual stream from a compact encoder pretrained on robot video with masked reconstruction, optical flow, and depth. The measured age |
21 pa...21 pages, 14 figures. Corresponding author: Lin Yang |
| Vote-in-Context: Turning VLMs into Zero-Shot Rank Fusers | 2026-10-31 | ShowIn the retrieval domain, candidates' fusion from heterogeneous retrievers is a long-standing challenge, particularly for complex, multi-modal data such as videos. While typical fusion techniques are training-free, they rely solely on rank or score signals, disregarding candidates' representations. This work introduces Vote-in-Context (ViC), a generalized, training-free framework that re-thinks list-wise reranking and fusion as a zero-shot reasoning task for a Vision-Language Model (VLM). The core insight is to serialize both content evidence and retriever metadata directly within the VLM's prompt, allowing the model to adaptively weigh retriever consensus against visual-linguistic content. We demonstrate the generality of this framework by applying it to the challenging domain of cross-modal video retrieval. To this end, we introduce the S-Grid, a compact serialization map that represents each video as an image grid, optionally paired with subtitles to enable list-wise reasoning over video candidates. ViC is evaluated both as a single-list reranker, where it dramatically improves the precision of individual retrievers, and as an ensemble fuser, where it consistently outperforms strong baselines like CombSUM. Across video retrieval benchmarks including ActivityNet and VATEX, the framework establishes new state-of-the-art zero-shot retrieval performance, demonstrating its effectiveness in handling complex visual and temporal signals alongside text. In zero-shot settings, ViC achieves Recall@1 scores of 87.1% (t2v) / 89.0% (v2t) on MSR-VTT and 99.6% (v2t) on VATEX, representing massive gains of up to +40 Recall@1 over previous state-of-the-art baselines. We present ViC as a simple, reproducible, and highly effective recipe for turning modern VLMs into powerful zero-shot rerankers and fusers. Code and resources are publicly available at: https://github.com/mohammad2012191/ViC |
Accep...Accepted to CVPR Findings 2026 |
| LocalNav: Distilling Frontier VLMs and Embodied RL for On-Device Object Goal Navigation | 2026-10-31 | ShowVision Language Models (VLMs) have emerged in the robotic domain as a powerful tool that enables environmental perception with language context, serving as a catalyst for open- vocabulary tasks like Object Goal Navigation (ObjectNav). Yet, their computational footprint typically confines them to cloud execution, hindering low-latency inference with local deployment on resource-constrained robots. To address this challenge, we present a distillation strategy that aligns a lightweight, 4B- parameter local VLM to our spatial-semantic navigation interface via frontier model traces for edge execution on embedded GPU devices (e.g., Jetson Orin). While cloud models establish a baseline achieving 41.61% Success Rate (SR) on the HM3D OVON benchmark, zero-shot deployment of our proposed architecture with Qwen3.5-4B reaches only 25.97% SR. We demonstrate that distilling frontier reasoning traces and subsequent finetuning with Embodied Reinforcement Learning from Verifiable Rewards (E- RLVR) obtains a 37.36% SR, substantially narrowing the gap to large cloud models. However, while deployable on memory- constrained devices, the model cannot compete with round-trip cloud inference time: we identify Token Generation (TG) as the main driver of latency and reduce it with Token Generation (TG)- regularized E-RLVR. The combined method reduces latency by 69.29% and retains a SR of 35.53%. Combined with quantization, this joint strategy yields a cumulative 80.21% reduction in overall inference latency without significantly sacrificing performance, presenting a viable paradigm for local, low-latency VLM execu- tion on mobile robots. |
|
| CheXPoT: Post-Hoc Visual Verification and Correction of VLM-Generated Radiology Reports | 2026-10-31 | ShowVision-language models (VLMs) have shown promise for chest radiograph report generation, but adoption remains limited because of hallucinated findings that are not grounded in the image. Correcting these hallucinations and improving overall report quality remains a significant challenge. Prior work has addressed this problem by refining models with reinforcement learning, preference optimization, or chain-of-thought supervision, often at substantial compute cost. To address this gap, we introduce CheXPoT, a program-of-thought (PoT) framework that verifies each generated finding against image features and removes those it cannot ground. CheXPoT decomposes generated reports into predicates with RadGraph and an LLM, grounds these predicates against a DenseNet-121 and Grad-CAM visual evidence table, executes PoT verification code, and rewrites reports after removing unsupported predicates. On a 4,145-image held-out split, CheXPoT yields statistically significant precision gains on 8 of 14 CheXpert classes and F1 gains on 5, with no significant decreases in either metric. CheXPoT lifts micro-average precision from 19.2% (MedGemma-4B) to 46.8% (MedGemma-4B + CheXPoT) and micro-average F1 from 23.0% to 49.4%, absolute gains of 27.6% and 26.4% respectively. We also tested CheXPoT on other VLM report generators. On LLaVA-Med, CheXPoT lifts micro-average precision from 4.5% to 17.7% and micro-average F1 from 6.3% to 19.4%. On MAIRA-2, CheXPoT lifts micro-average precision from 15.6% to 42.1% and micro-average F1 from 21.0% to 45.3%. Our results suggest that inexpensive post-hoc visual verification may improve VLM trustworthiness by correcting unsupported findings from generated reports. CheXPoT lays a foundation for tools that help validate VLM-generated radiology reports and reduce unsupported findings. These innovations may subsequently help increase trust and adoption of VLM tools in Medicine. |
15 pages, 12 figures |
| What Transfers from Text to Vision? Capability Scaling Laws and Transfer Dynamics for VLMs | 2026-10-31 | ShowChoosing the right large language model (LLM) backbone is the most consequential decision when building a vision-language model (VLM), yet it remains fundamentally unprincipled: compute-based scaling laws fail to generalize across model families, and no framework exists for directly predicting VLM performance before training begins. We propose the Capability-Driven Multimodal Scaling Law, the first cross-family framework that predicts VLM benchmark accuracy from directly observable textual capability. Given a low-dimensional capability score |
EMNLP 2026 Main |
| Spotter: Let the Embodied Model Lead, and the VLM Reflect for It | 2026-10-31 | ShowCurrent embodied models do not respond to their own failures, although what just went wrong could inform a small adjustment on the next attempt, the kind of reflection behind the gains of thinking in language models. We test whether they can repair a known error, which requires producing a correction and judging whether it is right. Stopped at a failure and allowed to retry, they seldom repair it through their own randomness or from a language description of the error, and best-of-N selection cannot pick the successful candidate after a failure. We attribute this to training only on successful demonstrations and to inputs too narrow to show what went wrong, and conclude that reflection must come from a vision-language model (VLM), which takes in far more information, such as the episode history and text, and is more general. Prior VLM-led work has the VLM plan every step and invoke the embodied model as a tool, placing the VLM on the critical path. We propose Spotter, which reverses the roles: the embodied model leads and executes continuously, while the VLM runs in parallel, monitors through a lightweight local screener, intervenes only when an error is detected, reflects on and corrects it, and returns control. We run Spotter with Qwen and with GPT as the VLM, and both improve the embodied models; with GPT, Spotter improves Cosmos Policy and |
19 pa...19 pages, 7 figures, 6 tables. Code: https://github.com/zqc3117/Spotter |
| PRISM-Net: Dynamic Routing Network with VLM-Guided Disentangled Attribute Distillation for Explainable Fine-Grained Product Retrieval | 2026-10-30 | ShowFine-grained visual product retrieval in large-scale e-commerce multimedia systems demands discriminative representations capable of simultaneously resolving macro-level category semantics and micro-level visual attributes such as structural contour (line) and chromatic appearance (color). However, conventional backbones including convolutional neural networks (CNNs) and vision Transformers (ViTs) inherently suffer from feature entanglement and shortcut learning. To learn the micro-level visual attributes without costly manual attribute annotations, we propose phase-routed interpretable semantic mixture-of-experts network (PRISM-Net), an explainable multi-expert architecture that disentangles holistic visual features into disentangled line and color expert pathways governed by a conditional meta-router. In phase 1, we introduce an external VLM-Guided sigmoid semantic distillation mechanism with gradient-detached routing gates, utilizing zero-shot text prompts from SigLIP to supervise specialized structural and chromatic experts. In phase 2, the disentangled expert embeddings are adaptively fused with macro base features via a learned meta-router gate to calibrate category boundaries. Extensive 5-fold cross-validation across 275,454 e-commerce images (18 categories) and full-scale retrieval benchmarks (55,091 queries against a 220,363-image gallery) across SwinV2, ConvNeXt-V2, and DINOv3 demonstrate substantial gains. On SwinV2, PRISM-Net boosts full-fold product mAP@10 from 80.98% to 83.90% (+2.92%), line mAP@10 by +5.74%, and color mAP@10 by +9.12%. Furthermore, unsupervised distribution analysis of the meta-router gate reveals that both ViT and CNN backbones emergently allocate up to 6.8x higher structural line weights to rigid appliances and hardware while dedicating >96% chromatic weight to clothing. |
10 pages |
| Why VLMs Miss Small Objects, and When Zooming In Is Safe | 2026-10-30 | ShowScreenshots, documents, engineering drawings, shelves and maps hold many small objects of kinds that change from task to task, and vision-language models (VLMs) promise to find them from an instruction and an example. Yet they see each image through a fixed token budget and miss small objects, so every user, and every agent that calls a model, must choose whether to send an image whole, raise its resolution, split it, or zoom. We show that VLMs miss small objects for two reasons a caller can estimate before any call: an object gets too few visual tokens (S), and a call must search and report too much (L). Raising S by a factor a multiplies the tokens needed to see an image by a^2, while splitting at the same scale lowers L at a cost at most linear in the number of views. Under two monotonicity assumptions, splitting cannot lower expected recall if no view zooms out and views overlap by one object, and neither part can be dropped; every such split pays a minimum token cost, and a simple rule pays it up to an explicit factor. Better models need less S, but an image can be arbitrarily large while a call's budget is fixed, so splitting does not become obsolete, for zoom agents either. In about 187,000 requests to 12 VLMs, up to GPT-6.1, Claude Opus 5.5 and Gemini 4, on controlled images 24 of 40 predicted orderings hold significantly and none is violated; splits that never zoom out raised recall on average over models wherever S was below 1.5, and splits that zoom out lowered it in 19 of 21 model and data settings. The rule raised recall significantly in five of ten settings, by up to 0.28 (from 0.62 to 0.90), and precision and F1 in six, at a token cost known before any call. |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| FloMo: 3D Scene Flow as a World Action Model Intermediate for Learning from Human Video | 2026-11-03 | ShowPretrained video generation models distill internet-scale motion priors into robot policies, but full video is a redundant signal for action prediction. We argue that dense 3D motion is the medium through which video pretraining best transfers to robot actions, and that predicting motion rather than video affords better sample-efficiency for action prediction. FloMo is a 5B video-generation model fine-tuned to predict 3D scene-flow and action tokens. Rendering scene flow as RGB and encoding it into the same latent space as video enables a pretrained video generation backbone to efficiently adapt to predict 3D scene flow. Co-trained on action-free human video and action-labeled robot teleoperation data, our model benefits from human-video pretraining and generalizes to unseen objects, motion primitives, and tasks. Replacing motion prediction with a video prediction objective degrades out-of-distribution generalization, while adding a video prediction target alongside motion lowers policy success rather than improving it. When motion, video, and action are predicted jointly, action tokens attend far more to motion than to video, suggesting that predicting video wastes model capacity. Together, these results indicate that motion effectively mediates the transfer from video pretraining to action. More information can be found at flomo-wam.github.io. |
|
| HiLo-WAM: Decoupling Long-Horizon Task Evolution and Local Dynamics in World Action Models | 2026-11-03 | ShowWorld Action Models (WAMs) improve robotic manipulation by jointly modeling visual dynamics and robot actions. However, most existing WAMs model dense short-horizon visual dynamics, capturing local interactions but providing limited long-range task context. Extending prediction horizons increases uncertainty and computation while entangling task-level evolution with fine-grained local dynamics, obscuring their dependencies and distinct roles in action generation. We propose HiLo-WAM, which decouples long-horizon task evolution from local visual dynamics while jointly modeling their interaction with robot actions. A sparse long-horizon visual stream captures historical task context and future evolution, while a dense local dynamics visual stream models short-horizon robot-object interactions. The two streams operate at different temporal scales within a unified dual-DiT architecture and guide actions through structured cross-stream attention. Progressive inference further reduces unnecessary computation through stream-wise partial denoising, joint denoising, and dynamic skipping. Experiments demonstrate strong long-horizon manipulation performance: HiLo-WAM achieves a 93.70% average success rate on RoboTwin 2.0 and a 17.43 progress score on RoboDojo long-horizon tasks, with robust in-domain and OOD performance on the Fourier GR-3 robot. Ablations confirm the complementary contributions of long-horizon task evolution and local interaction dynamics, while temporal-scale analyses highlight the importance of balancing temporal coverage and resolution. Progressive inference further achieves a 2.93x speedup while improving performance. |
|
| When Does Test-Time Search Help World Action Models? An Empirical Audit of Candidate Selection and Simulator Integrity | 2026-11-03 | ShowWorld-action models increasingly apply test-time scaling, spending extra inference computation on candidate selection and adaptive allocation to improve control. Such gains are credible only when candidates carry exploitable signal, nominal compute savings become real speedup, and evaluation stays reliable. We audit all three on Cosmos Policy and RoboCasa with a premise-first, pre-registered methodology. On clean data, a learned value selector is 1.66 pp worse than random; a hindsight oracle gains only 1.40 pp, with an episode-level interval spanning zero; and 82% of states show identical candidate outcomes, leaving little selection headroom. Token pruning cuts per-forward FLOPs by up to 47% but yields at most a measured 1.15$\times$ end-to-end speedup on the deployed path (1.25$\times$ on a decode-excluded path), short of a 1.5$\times$ target, and our adaptive-allocation experiment is inconclusive under a power gate. We further identify and fix a silent rendering corruption whose rate rises sharply along compute-heavy planning paths while leaving the official direct-policy result unaffected, and release a detector, fix, and scripts. These findings show why test-time scaling should be audited as a chain of premises (decision signal, realized efficiency, evaluationintegrity) rather than by success rate alone. Conclusions are scoped to the tested configuration. |
|
| Coordinating Humanoid Loco-Manipulation through Action-Conditioned Latent Foresight | 2026-11-03 | ShowHumanoid loco-manipulation requires coordinated body motion and hand interaction over time, and a plausible-looking command sequence can still drop a held object or leave an interaction incomplete. We present action-conditioned latent foresight, a latent world-action model (WAM) that evaluates proposed command sequences before execution. An action generator proposes joint motion-hand command sequences from vision, language, and the robot state. An action-conditioned endpoint predictor estimates the visual features each sequence prefix reaches and measures progress toward a demonstration-derived subgoal, while a joint temporal prior scores temporal consistency given the robot state and preceding commands. The robot executes the first commands of the best-scoring sequence and replans from the next observation. During training, action consistency regularization reconstructs the first demonstrated command from observed and predicted feature changes, encouraging the predictor to retain action-relevant information. On ten tasks in the humanoid loco-manipulation simulation benchmark SIMPLE, our method achieves 89.7% average success, compared with 73.7% for Psi-Zero and 82.0% for DiT4DiT. On seven real-robot tasks, it achieves 85.7% under standard conditions and 65.7% under unseen object configurations, while GR00T N1.7 achieves 70.0% and 48.6%, respectively. Project page: https://ju6276.github.io/Action-Grounded-Foresight/ |
10 pa...10 pages, including supplementary material |
| Cycle WAM: Coupling Action and Video through Cycle Consistency | 2026-11-03 | ShowRecent advances in video modeling have driven substantial progress in world action models (WAMs). Existing approaches exploit pretrained video priors by decoding actions from latent representations or inferring them from generated videos. The underlying link is causal: robot actions produce changes in the scene, and video observations of these changes provide evidence for inferring the actions that produced them. However, existing training formulations do not by themselves ensure that actions and their predicted visual consequences are consistent, potentially limiting the use of pretrained capabilities. We treat video and action as paired modalities linked by this causal relationship and learn their correspondence through a unified training framework designed to better exploit pretrained video priors. Cycle WAM first augments video-action co-training with bidirectional conversion to reinforce cross-modal correspondence and recover forward and inverse dynamics. It then introduces a differentiable cycle-reconstruction objective to further couple the two mappings. The resulting model supports both forward and inverse dynamics, with the potential to improve performance through more effective use of pretrained video priors. Cycle WAM achieves state-of-the-art performance among comparable methods on LIBERO-Pro and substantially improves DreamZero on DROID. Comparisons with UWM characterize its inverse-dynamics performance and show sharper visual detail in the displayed forward-dynamics predictions. |
|
| Future Representation Distillation for Imagination-Free World Action Models | 2026-11-03 | ShowWorld Action Models (WAMs) use the predictive knowledge of pretrained video diffusion models for robot control through future video prediction. However, iterative video denoising is costly, and inaccurate visual predictions can compromise action prediction. Our key idea is to make this predictive knowledge available through visual representations computed from the current observation, enabling action prediction without generating future videos. We introduce DreamFree, a WAM that repurposes a pretrained video diffusion model as a predictive visual encoder through representation distillation. The visual encoder learns to predict a frozen video diffusion model's representation using only the current observation. An action module is then trained to predict actions by extracting action-relevant information from the visual encoder. Across real-world and simulated tasks, DreamFree achieves higher success rates than state-of-the-art VLAs and WAMs, while performing real-time closed-loop control at 11Hz with a 14B video diffusion backbone. Experiments show that DreamFree can translate stronger video priors and additional video data into better robot control and enable learning new robot tasks from video demonstrations, including videos from a different embodiment, without action annotation for the new tasks. |
|
| Seeing Through the Illusion: Test-Time Scaling for World Action Models | 2026-11-02 | ShowWorld action models (WAMs) jointly generate actions and imagined futures, but a plausible prediction need not be realized by its co-generated action. We introduce WISE (\emph{World Action Inference Scaling through Evaluation}), a pre-execution selector that scores each sampled rollout on action validity and temporal regularity, inverse-dynamics consistency between the proposed action and the imagined motion, and instruction-conditioned task progress, and executes the highest-scoring candidate while keeping the WAM frozen. With four candidates per decision, WISE raises closed-loop success from 6.0% to 24.0% on RoboLab with Cosmos~3 Edge and from 75.1% to 78.3% mean success on LIBERO with DreamZero, and exceeds the best existing selector in aggregate by 4.0 and 0.9 points. Component ablations and agreement analyses show that the three scores rank candidates differently and that task progress carries the largest weight in the fused decision. |
Accep...Accepted at the CoRL 2026 Workshop on From Generative Models to Robot Actions (G2A) |
| TurboWAM: Learning What Matters for Fast World Action Models | 2026-11-02 | ShowWorld action models have demonstrated promising performance in robot policy learning, yet their high inference latency remains a critical bottleneck for real-world deployment. Recent approaches have sought to improve inference efficiency by eliminating test-time future-frame imagination or accelerating action prediction. However, aggregating representations from all video world model layers can introduce redundant computation, while standard backpropagation may distribute action-supervision gradients unevenly across layers, limiting the effectiveness of world representation learning. In this work, we propose TurboWAM, a fast world action model that learns to efficiently aggregate and optimize world representations for action prediction. Specifically, TurboWAM introduces an efficient world representation aggregation mechanism that restricts aggregation to a bounded range of relevant world model layers identified through layer relevance analysis. Within this bounded range, TurboWAM estimates aggregation weights for different world model layers based on the current action denoising state and combines them with learnable layer-wise weights to fuse world representations. Moreover, TurboWAM employs a world model gradient modulation mechanism that decomposes action-supervision gradients into a layer-shared component and layer-specific residual components, enabling more balanced gradient propagation across world model layers. This mechanism operates exclusively during training and introduces no additional inference computation. Together, these designs enable more efficient world representation learning and low-latency action prediction. Extensive experiments on multiple robot manipulation benchmarks and real-world manipulation tasks show that TurboWAM achieves low inference latency while maintaining competitive task success rates compared with state-of-the-art methods. |
Techn...Technical report. Project page: https://turbowam.github.io/ |
| SV-WAM: An Efficient Surround-View World-Action Model for End-to-End Autonomous Driving | 2026-11-02 | ShowWorld models (WMs) have demonstrated strong potential for end-to-end autonomous driving by learning predictive representations of future scene dynamics. However, generating future videos during inference introduces substantial computational overhead, leading many recent driving WMs to adopt a single front camera as input for efficient deployment. This design restricts spatial coverage in safety-critical maneuvers such as lane changes, merges, and turns. To address this limitation, we propose SV-WAM, a surround-view world-action model (WAM) that preserves full six-camera observations while maintaining efficient inference. SV-WAM leverages future-video prediction as dense training supervision for action learning within a shared generative model, rather than as an inference-time output. At the core of this design is an action-centered causal mask that prevents action tokens from attending to future-video tokens during joint action-video denoising. Consequently, the video branch can be discarded at deployment, enabling efficient action-only planning. Furthermore, we introduce a differentiable drivable-area compliance regularizer that penalizes vehicle-footprint corners approaching or crossing drivable boundaries, improving planning safety and boundary awareness. Extensive experiments on the closed-loop NAVSIMv2 benchmark and the open-loop nuScenes benchmark demonstrate that SV-WAM achieves state-of-the-art planning performance with low inference latency and competitive zero-shot transfer capability. |
23 pa...23 pages, 16 figures. Project page: https://jinyangwang427.github.io/SV-WAM/ |
| RoboSeed: Multidimensional Evaluation of Vision-Language-Action and World-Action Models: Capability, Resource Affordability, and Research Evidence | 2026-11-02 | ShowChoosing a robot action model requires evidence about task capability, adaptation cost, deployment requirements, and the reliability of reported conclusions. Success rates alone offer limited guidance to researchers with different computing budgets and goals. We propose RoboSeed, a multidimensional evaluation framework for vision-language-action (VLA) and world-action models (WAM), within a continuing program to train, evaluate, and openly release robot models with verifiable evidence for model selection. The framework defines five dimensions: capability, downstream training affordability, deployment affordability, locally verified research value, and reproduction evidence. It maps empirical measurements and audit criteria to 0-100 profiles using fixed normalization anchors, with balanced, training-constrained, and deployment-constrained preference settings. Interface and protocol checks establish when comparisons are appropriate. Missing measurements remain explicit, incomplete aggregate scores are withheld, and weighted input coverage identifies evidence gaps. We apply the framework to representative VLA and WAM models, drawing on local training, released-weight evaluation, task-level audits, and an evaluation-system study. The results reveal budget-dependent backbone choices, sensitivity to adaptation recipes, and differences between language and visual task interfaces. Reproducible scoring code and machine-readable evidence support extension to additional models. The study provides dimension-level assessments and priorities for further measurement, while incomplete serving traces and heterogeneous protocols preclude overall numerical rankings. |
Proje...Project repository: https://github.com/PrimitiveX/RoboSeed |
| W2A: Distilling World Priors into Action Models for Real-Time Robotic Control | 2026-11-02 | ShowWorld Action Models (WAMs) have recently shown great promise in robotic control, benefiting from the future imagination capability of large-scale pretrained video models. However, most existing WAMs require explicitly imagining future states at test time, incurring substantial computational overhead and latency that make them unsuitable for real-time robotic control. In this paper, we propose W2A, a two-stage distillation framework for real-time robotic manipulation that distills intermediate denoising representations from a video diffusion model into an action model during training, while entirely discarding the video model at test time. We hypothesize that although video latent representations are high-dimensional, a compressed low-dimensional subspace is sufficient to provide effective supervision for action generation. This makes it possible to transfer such information effectively into an action model. Extensive experiments in both simulation and real-world settings demonstrate competitive success rates while significantly improving inference efficiency. On LIBERO, W2A achieves a 95.9% average success rate. More importantly, W2A achieves about 21x faster inference than the representative WAM baseline. |
|
| Gnomon4D: Learning Where and When to Focus for Geometric Motion-Grounded World-Action Modeling | 2026-11-02 | ShowWorld-action models (WAMs) use video-generation priors to jointly predict future observations and robot actions. Yet RGB futures entangle physical state changes with appearance, while frame-wise depth and point maps lack temporal correspondence. Existing 4D approaches either keep motion implicit in visual features or translate predicted traces into actions through a separate interface. We present Gnomon4D, a WAM that jointly generates future video, robot actions, and GeoDelta, a video-aligned 4D geometric motion field encoding the cross-frame image-plane displacement and depth change of embodiment and object points. By placing embodiment motion and object responses in a shared space, GeoDelta exposes their motion relations to the action expert at inference. Interaction-Salient Attention (InSAttn) uses these relations to focus spatial readout on task-relevant interaction regions, while Transition-Anchored Aggregation (TransAgg) aggregates evidence around changes in their relative motion, such as grasp and release. GeoDelta describes surface motion independently of embodiment-specific controls, providing a shared human--robot quasi-action for action-free pretraining on EgoManip-5000H followed by robot mid-training on Robot-30H. Gnomon4D leads on RoboTwin~2.0, VLABench, and RoboDojo, and outperforms the strongest real-robot baseline by 7.0 and 16.0 percentage points under in-distribution and out-of-distribution settings, respectively. |
|
| LOGIV: Graph-Verified Execution with Causal Local Repair for Long-Horizon Robot Tasks | 2026-11-02 | ShowAgentic robots must turn a high-level goal into actions while keeping the physical conditions needed for future steps valid. A completed action does not ensure that its effects will persist: a later disturbance can invalidate support that the remaining plan still assumes. We present LOGIV, a graph engineering framework for agentic robot execution that uses an explicit task graph to connect symbolic planning, observation-based verification, and embodied action. LOGIV compiles a validated plan into condition-support and resource-ordering edges, checks their physical validity during execution, and uses dependency tracing to determine what must be restored after a failure. Repair is constrained by retained unfinished actions, and the combined continuation is symbolically validated before execution resumes. The framework operates above existing Vision-Language-Action (VLA) and World Action Model (WAM) policies without adding trainable components. Across 20 dual-arm tasks from RoboTwin V2 and BiCoord with three execution backbones, LOGIV improves mean task success without injected disturbances by 10.67 and 15.33 percentage points, respectively. Under predefined perturbations, it achieves higher task completion and condition restoration than three adapted recovery baselines for every backbone on both benchmarks. On three real-robot tasks, successful trials increase from 12/30 to 17/30. These results support explicit, continuously checked task dependencies as a practical interface for reliable agentic robot execution. |
14 pa...14 pages. Updated related work, references, author list and affiliations |
| ODE Policy: Robot Learning via Action Dynamics, Not Actions | 2026-11-01 | ShowMost robot policies, whether trained by imitation learning (IL) or reinforcement learning (RL), map observations to action values at discrete control steps and leave the continuity of physical motion to be learned implicitly. We propose ODE policy, which learns the dynamics of the action rather than the action itself. A neural ordinary differential equation (ODE) learns the vector field of the action by matching the integrated actions to the targets over short action chunks in IL and single control steps in RL. A lightweight diversity head captures multi-modal behavior by randomizing the direction of the field. At deployment, each action is integrated from the previous one, so the executed trajectory is continuous by construction. We evaluate it as a visuomotor policy, as the action head of vision-language-action (VLA) and world action models (WAMs), and as the actor of model-based RL against 14 baselines on unimanual, bimanual, dexterous, locomotion, and loco-manipulation tasks. In IL, against flow matching, it reaches 5.9$\times$ the success, 5.4$\times$ the generalization, and 3.9$\times$ the mode balance after one epoch, and generates actions 10.8$\times$ faster. It mostly moves more smoothly than human demonstrations. On 14 real-world tasks across single-arm, dual-arm, dexterous, and humanoid robots, it achieves the highest success rate. In RL, it further speeds up SuperMimic for humanoid motion tracking and smooths its actions. Efficient in learning, data, parameters, and execution, ODE policy offers a promising path to scaling robot learning. |
35 pages, 23 figures |
| Xiaomi Auto World Model 2.0: Causal World Models for Real-World Driving and Embodied Intelligence | 2026-11-01 | ShowWe present Xiaomi Auto World Model 2.0, a unified streaming causal world model for Physical AI. It models and generates causal physical evolution in the real world, and unifies heterogeneous conditioning inputs for view-consistent generation. The architecture integrates View-Temporal Attention (VTA) to capture cross-view spatial dependencies and temporal dynamics, together with Multi-view Rotary Position Encoding (RoPE) to encode positional relationships across views and time. WorldRec 2.0 reconstructs static and dynamic 3D Gaussians from multi-view, multi-temporal observations in a single forward pass and uses their renderings as explicit scene context to preserve geometry and appearance during long-horizon generation. Conditioning inputs are fused through VAE latent concatenation in a shared representation space. We incorporate Gated Linear Attention (GLA) into the temporal layers of VTA for efficient streaming inference, reducing computation and KV-cache memory during multi-view rollouts. A physical-interaction data pipeline covers diverse interactions and long-tailed events. From Pre-train onward, these data are combined with large-scale videos and heterogeneous visual modalities to learn visual, geometric, and dynamic priors. During Mid-train, training across diverse view configurations further improves the model's ability to capture how actions and object states shape subsequent scene evolution. After distillation-based Post-train, we apply the model to closed-loop simulation, closed-loop training, and World Action Model (WAM) applications. These applications generate physically consistent future states and support world-state prediction, policy learning, and action generation for Physical AI. |
40 pa...40 pages, 20 figures. Project page: https://xiaomiautoworldmodel.github.io/CausalWorldModel/ |
| Adapting Video Models for Robot Control via Motion-guided Asymmetric Prediction | 2026-11-01 | ShowWorld Action Models (WAMs) leverage pretrained video generative models for robotic control by jointly predicting future visual observations and actions. The visual prediction objective plays a key role in adapting pretrained video representations to the target robot domain. Yet standard visual prediction inherited from video pretraining treats all visual regions uniformly, despite their differing relevance to robotic control. We therefore investigate the spatial allocation of visual supervision in WAMs and show that focusing supervision on dynamic regions can substantially improve policy performance. Building on this finding, we introduce Motion-Guided Asymmetric Prediction (MoGAP), which augments WAM training with motion-focused visual prediction. MoGAP allocates noise asymmetrically across regions, training the model to reconstruct more heavily corrupted dynamic regions using less corrupted static regions as context. MoGAP is trained alongside standard video-action prediction through a training-only auxiliary prediction path within the shared video diffusion backbone. It uses motion cues derived directly from training videos, requiring neither an external teacher nor additional modalities. Across robot manipulation benchmarks and WAM architectures, MoGAP consistently improves policy performance, increasing success rates on the GR-1 tabletop benchmark from 38.8% to 47.5% for Fast-WAM (Wan2.2-5B) and from 52.3% to 62.3% for Cosmos3-Nano. |
Proje...Project page: https://ismty0805.github.io/mogap/ |
| pico-WAM: Evaluating Counterfactual Prediction and Planning Error in a Compact World Action Model | 2026-10-31 | ShowWhen a World Action Model fails at control, is the bottleneck proposal generation or trajectory scoring? We introduce pico-WAM, a compact model with supplied physical state for tracing policy-dynamics interactions. Using 206 Push-T demonstrations and 15 training seeds, we compare data regimes and search configurations. Counterfactual branch training improves short-horizon cost ranking and enlarges the coverage advantage of policy-restarted over warm-started search by 4.8 percentage points; restart wins under both data regimes. Replacing learned trajectory scores with simulator costs improves coverage by 12.2 points while retaining learned proposals. Substituting simulator states in proposal generation alone shows no detected gain. Component diagnostics and a separate door-opening study show why better ranking need not yield better control or outperform imitation. These controlled interventions identify scoring as an effective improvement target in the tested setting and provide a reproducible protocol for diagnosing compact world models. |
12 pa...12 pages, 8 figures. Code available at https://github.com/keivalya/pico-wam |
| CausalWAM: Learning the Causes of Actions in World Action Models | 2026-10-30 | ShowWorld action models (WAMs) learn latent variables from observations and use these to both predict the evolution of the scene and to generate actions. The prediction has to be of the whole scene, whereas only a portion of the scene causes actions. WAMs fitted to demonstrations do not, in general, guarantee that they will generate actions solely from the latent causes of the demonstrated actions (the action-relevant latent variables), as the model could just as well learn to imitate in this way whilst entangling the action-relevant latent variables with the nuisance variables. We therefore separate out what causes action from what prediction depends on. We prove that under a sparsity constraint the latent variables used to generate actions are an invertible function of the action-relevant latent variables alone and the true structure is identified up to a permutation of the latent variables. This requires neither sufficient temporal dependence nor any assumptions on how the unobserved regimes evolve and holds even when the action-relevant latent variables are statistically dependent on the rest. We therefore propose CWAM (Causal WAM), which incorporates the constraint into a pretrained WAM by adding a sparse routing path to its action head, with the world modeling objective remaining unchanged. Experiments on both synthetic data and real-world robots validate the theory as well as showcase the improved efficiency in closed-loop success. |
|
| When the Simulator Fails, Not the Policy: Auditing Contact Artifacts in Simulated Robot Benchmarks | 2026-10-30 | ShowSimulation benchmarks score robot policies by success rate, and every failed episode counts against the policy. Many of these failures are the simulator's fault, not the policy's; we find the cause and fix it. Testing a vision language action model and a world action model in simulation, we find that the policy often places the object and the simulator then ejects it, and that the benchmark's own scripted demonstrations fail the same way. We present SimuGuard, an auditing plugin for LIBERO, RoboTwin, RoboCasa and RoboDojo. It leaves the evaluation unchanged. At every physics step, it records the state and flags an object that starts moving at a contact faster than the robot's contact or gravity account for. It replays a recorded episode bit for bit, and it reruns the episode with one simulation parameter changed. With it we trace these episodes on RoboTwin to the collision detection of the physics engine: for geometry one to two centimeters apart, the engine reports a penetration of about two centimeters, and the solver throws the object out at several meters per second. The cause is one constant: the distance algorithm discards triangles below a fixed area threshold, which valid triangles fall below at the centimeter scale. With that constant changed and nothing else, episodes with an artifact fall from 12 in 100 to none, and the success rate on five container tasks rises from 89% to 98%. Success rates in simulation thus count failures that the simulator causes. Benchmarks on PhysX 5.3.1 or earlier should rerun in a repaired engine. Where the engine cannot change, they should report the success rate together with the range it spans when every flagged episode fails or passes: 75.8% to 98.5% around a reported 83.1% for LingBot-VA. |
|
| Adaptive-WAM: Quality-Guided Early-Exit Planning from Intermediate Video-Diffusion Features | 2026-10-30 | ShowLarge video diffusion models provide rich spatiotemporal priors for autonomous driving, but existing world-action models often inherit the cost of iterative future-video generation even though deployment only requires an ego trajectory. We ask a more basic question: how much of a video diffusion model must be executed to make a reliable driving decision? Through a controlled study of video denoising timesteps and Diffusion Transformer (DiT) depth, we find that planning performance is largely insensitive to the tested video-noise levels, whereas strong trajectories can already be decoded from intermediate layers. Based on this observation, we introduce Adaptive-WAM, a quality-aware multi-exit planner built on a Wan2.2-5B backbone. Trajectory diffusion heads are attached to selected DiT blocks, and a lightweight trajectory-quality scorer terminates inference once the best trajectory decoded so far satisfies a quality threshold; otherwise, computation continues from the cached hidden state to a deeper exit. The deployed planner therefore avoids the iterative classifier-free denoising loop and VAE decoding required for future-video synthesis, while dynamically allocating backbone depth according to trajectory quality. On NAVSIM, the adaptive single-trajectory planner achieves 90.8 PDMS; a separate fixed-exit variant reaches 92.6 PDMS with 64 proposals. It further obtains 89.9 EPDMS on NAVSIM v2, yielding the best reported results among the compared front-view video world-model planners. Without target-domain fine-tuning, Adaptive-WAM transfers to nuScenes with 0.88 m average L2 error and a 0.08% collision rate. On an A100, adaptive routing improves PDMS from 90.62 to 90.79 while averaging 170 ms end-to-end planning latency, approximately 10% below the 190 ms fixed block-15 planner and 47% below the 320 ms fixed full-depth planner. Code will be released. |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| GSCV++: Lightweight 3D Gaussian Splatting Sequence Compression via Standard Video Codecs | 2026-11-03 | ShowThis paper presents GSCV++, a lightweight pipeline for 3D {G}aussian {S}platting (GS) sequences {C}ompression with standard {V}ideo codecs. To meet the requirements of low-complexity and hardware compatibility from the industry, GSCV++ jointly improves preprocessing, 3D-to-2D mapping, and practical video compression configuration of GS compression. First, for preprocessing, energy-adaptive inter-frame spherical harmonic (SH) alternative current (AC) dimensionality reduction (EA-SHDR) is proposed to produces a compact and stable temporal SH AC feature space. It can reduce the data volume of following mapping and practical compression, as well as less complexity while eliminating redundant information. Second, for 3D-to-2D mapping, prior-based Inter-MiniPLAS (P-MiniPLAS) is designed to generate smooth 2D feature maps through prior-based weighting factor inference and small scale remapping. Compared with vanilla Parallel Linear Assignment Sorting (PLAS) and Morton scan, P-MiniPLAS ensures the generated 2D maps are more suitable for the video codec while maintaining a low level of computational complexity. Third, for practical compression, content-aware generalized compression (CA-GC) is developed to use a non-uniform bit-depth split for coordinate map generation, and a coding unit size selection is introduced for efficient compression complexity control, which fully exploits the performance of video codecs without involving normative changes. Experiments demonstrate that GSCV++ reports over 20% in intra- and 60% in inter-prediction rate-distortion gains compared with state-of-the-art (SOTA) methods. GSCV++ also achieves faster 2D map generation, encoding, and decoding than other SOTA methods using the same type of video codecs. Ablation studies verify the effectiveness and robustness of the proposed modules and provide insights for subsequent research. |
|
| GSPIM: A Temporal-Locality-Aware 4DGS Accelerator for PIM-Enabled Edge Devices | 2026-11-03 | Show4D Gaussian Splatting (4DGS) enables high-quality real-time novel-view synthesis for dynamic scenes, supporting free-viewpoint video, mobile XR, and autonomous systems at the edge. Methods using explicit 4D distributions, keyframe interpolation, or anchor binding share two GPU data-movement bottlenecks. Temporal filtering follows model record loading, wasting LPDDR5 bandwidth on inactive data. Frame-isolated execution also overlooks highly overlapping active sets and depth relationships across adjacent frames, reloading shared attributes when working sets exceed on-chip cache capacity. To address these bottlenecks, we present GSPIM, a temporal-locality-aware 4DGS accelerator for PIM-enabled edge devices. Polymorphic PIM adapts a shared near-memory datapath to representation-specific field layouts and selection procedures. It filters and compacts valid data into contiguous inputs before the GPU accesses complete attributes, reducing ineffective transfers. Multi-Frame Rendering Dataflow reuses active attributes and stable depth relationships within a temporal window, reducing repeated work across frames. Hybrid Dataflow Scheduling overlaps near-memory data preparation with GPU rendering to reduce waiting. Evaluation combines a GPU-PIM model calibrated to Jetson AGX Orin 64 GB with a 1z DRAM physical implementation of the added circuits. Across three 4DGS methods and six dynamic scenes, GSPIM achieves a 4.15x geometric-mean speedup and a 4.99x geometric-mean energy-efficiency improvement over a fully optimized frame-by-frame GPU baseline. The maximum scene-level PSNR difference is 0.011 dB, and the added physical footprint is 18.7172 square millimeters across 32 LPDDR5 dies. |
15 pa...15 pages, 23 figures, Accepted by HPCA 2027. Code repo: https://github.com/MAdrid1011/GSPIM |
| Phase-DGS: Phase-guided Dynamic Gaussian Splatting from Unsynchronized Multi-view Video | 2026-11-03 | ShowSpatiotemporal scene reconstruction with Dynamic Gaussian Splatting (DGS) fundamentally depends on perfect temporal synchronization across all cameras, an assumption rarely achievable in practice. Industry-standard synchronization requires expensive specialized hardware and complex setup, while data-driven alternatives including audio-based alignment, geometry matching, and pose tracking remain limited by restrictive environmental or scene-specific assumptions. To overcome these limitations, we propose Phase-DGS, a framework that replaces conventional timestamps with a semantic phase representing each frame's inherent state within the underlying motion cycle. Frames capturing the same motion state share the same phase regardless of which camera records them or when, naturally establishing spatiotemporal correspondence across unsynchronized cameras. We validate Phase-DGS under temporal offsets, frame drops, and camera freezes, achieving up to +8.5 dB PSNR improvement over baselines while enabling seamless integration with existing DGS backbones including RealTime-4DGS and FreeTimeGS. |
Accep...Accepted to NeurIPS 2026. Project page: https://oc-dev74.github.io/Phase-DGS/ |
| EmbedTalk: Talking Head Synthesis using Gaussian Embeddings | 2026-11-03 | ShowDeformable 3D Gaussian Splatting (3DGS) has emerged as a popular method for real-time talking head synthesis, offering high-quality renderings at low latency. Tri-planes are a common choice for encoding Gaussians prior to deformation since they provide a compact and continuous representation. However, tri-plane encodings are limited by grid resolution and approximation errors introduced by projecting 3D volumetric fields onto 2D subspaces. Recent work has demonstrated the effectiveness of per-Gaussian embeddings for driving temporal deformations in 4D scene reconstruction. We introduce EmbedTalk, which leverages these embeddings to model speech-driven facial deformations for talking head synthesis. Comprehensive experiments show that EmbedTalk improves rendering quality, lip synchronisation, and motion consistency over previous 3DGS-based methods, while remaining competitive with state-of-the-art generative models. Replacing tri-plane features with embeddings also yields significantly more compact models that achieve 60+ FPS on a laptop GPU (RTX 2060 6 GB). Code and further details are available at https://arpita2512.github.io/EmbedTalk/ |
Camer...Camera Ready version for WACV 2027 |
| 3DGS-VBench+ and 3DGSV-Assessor: A Multi-Dimensional Subjective Benchmark and a Reference-Flexible Objective Model for 3D Gaussian Splatting Video Quality Assessment | 2026-11-03 | Show3D Gaussian Splatting (3DGS) enables real-time novel view synthesis (NVS), but its heavy storage makes compression essential, and both training and compression introduce representation-specific distortions. Since generative compression in many methods treats geometric and color attributes via different parameters, it can induce dimension-specific distortions, motivating multi-dimensional assessment. Moreover, real-world captures generally lack an aligned ground-truth video for a synthesized, continuous camera trajectory, leaving only sparse captured views as references. To enable multi-dimensional quality assessment of 3DGS videos with scarce aligned references, we first present 3DGS-VBench+, a large-scale subjective video quality assessment (VQA) benchmark comprising 3{,}036 3DGS reconstructions, each paired with a rendered video, across 11 real-world and 22 synthetic scenes, generated by 6 representative compression algorithms at systematically designed levels and annotated with 9{,}108 MOSs covering overall, geometry, and color quality. Based on 3DGS-VBench+, we propose 3DGSV-Assessor, a reference-flexible, large multimodal framework that scores all three dimensions in one generation pass and fuses a ranking-based reference cue into the frozen regressor through a bounded gate, supporting both reduced-reference (RR) and no-reference (NR) assessment. Extensive experiments demonstrate state-of-the-art correlation with human perception on 3DGS-VBench+ and strong zero-shot generalization to other NVS benchmarks. The dataset and code will be released at https://github.com/YukeXing/3DGSV-Assessor. |
|
| Physical Diffusion Policy via MPM-Guided 3D Gaussian Simulation | 2026-11-03 | ShowLearning robust manipulation policies for diverse everyday objects remains challenging because objects with different materials exhibit distinct deformation and contact dynamics. While diffusion policies provide a powerful framework for modeling multimodal manipulation behaviors, they typically lack explicit physical reasoning and may produce averaged actions when facing heterogeneous objects. We present a physics-aware diffusion policy that augments visual imitation learning with object physical understanding. Our framework recovers material properties from visual demonstrations by coupling 3D Gaussian Splatting with differentiable Material Point Method simulation, where video guidance and particle-level constraints enable accurate physical parameter estimation. The estimated physics conditions are further integrated into a material-aware Mixture-of-Experts diffusion policy, allowing adaptive behavior generation across different physical regimes. For practical deployment, we introduce a physics memory bank that provides efficient physical retrieval without requiring online simulation optimization. Experiments on a wide range of simulated and real-world objects demonstrate improved manipulation success and generalization over existing diffusion-based approaches. These results show that incorporating physical knowledge is critical for scalable robot manipulation in diverse environments. |
10 pages |
| IGSW: Multi-Weather Urban Scene Reconstruction with Intrinsic-Decoupled Gaussian Splatting | 2026-11-03 | ShowHigh-fidelity weather editing of 4D scenes can support end-to-end training and closed-loop simulation for autonomous driving. Combining generative image editing with dynamic Gaussian representations enables multi-weather scene reconstruction, but weather-edited images often contain structural distortions. Directly fitting their appearance may propagate errors from teacher images into the reconstructed 4D scene, compromising the original scene structure and temporal continuity. To address these challenges, we propose IGSW, a two-stage framework that combines intrinsic-decoupled 4D Gaussian scene modeling with structure-preserving weather refinement. In Stage I, background Gaussian appearance is decomposed into weather-shared albedo and normals and weather-specific irradiance and residuals. Weather teacher images generated by a diffusion-based image editing model and intrinsic maps estimated by a diffusion-based inverse renderer provide feature-level soft supervision for stable structure and illumination changes. In Stage II, a structure-preservation gate combines semantic conflicts with cross-weather intrinsic discrepancies to suppress unreliable teacher regions. Per-Gaussian weather transport variables and target-weather branches then strengthen the desired weather appearance over the intrinsic-decoupled scene. Experiments on nuScenes and Waymo show that IGSW achieves the highest content similarity and semantic consistency and the lowest temporal error among the compared methods, while retaining pronounced weather-dependent illumination and atmosphere. The project page is available at https://akaxi6.github.io/IGSW/ . |
|
| EndoS3:Endoscopic Semantic Splatting SLAM | 2026-11-03 | ShowPurpose: Despite advances in 3D surgical scene understanding, semantic simultaneous localisation and mapping (SLAM) remains under-explored for two key reasons. First, it requires solving three challenging tasks in parallel: camera pose tracking, high-fidelity deformable reconstruction, and semantic mapping, making progress particularly demanding. Second, the scarcity of datasets providing the necessary ground truth to evaluate all three tasks comprehensively under conditions that closely reflect real surgical procedures environments. Method: We propose EndoS3, a Gaussian Splatting-based semantic SLAM framework that uses instance masks to guide all three tasks. An instance-level motion decomposition module separates static and dynamic instances by evaluating residual flow after compensating for camera motion, accounting for flow noise and comparing motion across neighbouring instances. The camera is tracked against the static Gaussians, while an instance-aware warp field gives each dynamic instance its own control nodes and deformation network. A semantic node allocation module redistributes a fixed node budget across instances according to their size and non-rigidity. We also introduce CRCD++, an ex-vivo Cholecystectomy dataset with RGB-D frames, camera pose and instance annotations for full semantic SLAM evaluation. Results: EndoS3 surpasses all state-of-the-art general-purpose and surgical semantic SLAM methods in tracking, reconstruction and semantic mapping on the CRCD++ and SemSup datasets. Conclusion: Semantic-aware motion and deformation modelling is vital to surgical semantic SLAM. EndoS3 and CRCD++ offer a method and benchmark for improved geometric and semantic understanding in endoscopy. |
|
| Unified Loss-Aware Density Control for 3D Gaussian Splatting | 2026-11-03 | ShowThe efficiency and fidelity of 3D Gaussian Splatting critically depend on adaptive density control, which grows and suppresses Gaussian primitives throughout optimization. However, the standard pipeline still relies on heuristic view-space gradients, scale thresholds, and global opacity resets, rather than explicitly assessing the loss-reducing effect of each split, clone, or opacity update. This mismatch can allocate primitives to already-sufficient regions, miss under-reconstructed structures, and destabilize training through blanket opacity changes. In this work, we recast adaptive density control as a collection of local loss-reduction decisions. We derive efficient Taylor-based per-Gaussian surrogates that predict the objective change caused by three primitive-level operations: splitting, cloning, and opacity decay. This yields a unified loss-aware framework with three components. First, we extend second-order splitting to a normalized joint position-scale space, enabling children to adapt not only their locations but also their anisotropic extents. Second, we formulate duplicate cloning as a surrogate minimization problem, selecting only Gaussians whose duplicated contribution is predicted to improve reconstruction and assigning each child a principled opacity. Third, we replace global opacity reset with selective soft opacity decay that weakens harmful primitives while preserving useful ones. Across standard 3DGS benchmarks, our method produces more targeted primitive allocation and achieves improved rendering quality with comparable or smaller Gaussian budgets compared to baselines. |
Accep...Accepted to NeurIPS 2026 |
| AniGS: Bridging Rendering and Diffusion Prior for 3D Scene Animation | 2026-11-03 | ShowNovel view rendering of large and complex reconstructed scenes is becoming increasingly photorealistic. However, most reconstructions remain static and lack the ambient motion that makes environments immersive. We present AniGS, a method for scene-level animation of 3D Gaussian Splatting (3DGS) reconstructions that adds subtle, distributed dynamics, e.g., vegetation motion, while preserving rigid structures. Unlike existing 3D animation techniques which are limited to object-centric subjects or small regions, AniGS is designed for large, cluttered, navigable scenes. AniGS represents the scene with a canonical 3DGS and models motion using a time-conditioned deformation field. To animate the entire scene, we leverage a pretrained video diffusion model and introduce an iterative dataset--model update strategy that progressively expands viewpoint coverage and repeatedly updates camera-fixed training videos using a render-and-refine scheme. To prevent artifacts from unintended motion in static areas, we further introduce a composed video-to-video refinement scheme that restricts motion to desired regions. Experiments on five real-world, large-scale outdoor scenes demonstrate that AniGS produces natural ambient dynamics and high-quality novel view videos, enabling more immersive viewing experiences of reconstructed environments. |
Prepr...Preprint. Project page: https://yccyenchicheng.github.io/AniGS/ |
| Aes3D: Aesthetic Assessment in 3D Gaussian Splatting | 2026-11-02 | ShowAs 3D Gaussian Splatting (3DGS) gains attention in immersive media and digital content creation, assessing the aesthetics of 3D scenes becomes important in helping creators build more visually compelling 3D content. However, existing evaluation methods for 3D scenes primarily emphasize reconstruction fidelity and perceptual realism, largely overlooking higher-level aesthetic attributes such as composition, harmony, and visual appeal. This limitation comes from two key challenges: (1) the absence of general 3DGS datasets with aesthetic annotations, and (2) the intrinsic nature of 3DGS as a low-level primitive representation, which makes it difficult to capture high-level aesthetic features. To address these challenges, we propose Aes3D, the first systematic framework for assessing the aesthetics of 3D neural rendering scenes. Aes3D includes Aesthetic3D, the first dataset dedicated to 3D scene aesthetic assessment, built on our proposed annotation strategy for 3D scene aesthetics. In addition, we present Aes3DGSNet, a lightweight model that directly predicts scene-level aesthetic scores from 3DGS representations. Notably, our model operates solely on 3D Gaussian primitives, eliminating the need for rendering multi-view images and thus reducing computational cost and hardware requirements. Through aesthetics-supervised learning on multi-view 3DGS scene representations, Aes3DGSNet effectively captures high-level aesthetic cues and accurately regresses aesthetic scores. Experimental results demonstrate that our approach achieves strong performance while maintaining a lightweight design, establishing a new benchmark for 3D scene aesthetic assessment. |
|
| 3D Gaussian Splatting within Self-Learned Neural View-Dependent Color Fields | 2026-11-02 | Show3D Gaussian Splatting (3DGS) has become a vital tool in novel view synthesis. 3D Gaussian functions with attributes like color, opacity, and scale are stacked together to explicitly model a radiance field. However, discrete 3D Gaussians usually produce poor depth with few geometry details but lots of floaters, which hinders the effective modeling of smooth radiance fields. To address this issue, we propose to learn a radiance field within a continuous view-dependent color field parameterized by neural networks. With continuous constraints in color, the bias on low-frequency signals of neural networks can highly encourage the geometry modeling to contribute to high-frequency color variations on images, getting rid of solely relying on the color attribute itself. Moreover, we impose a constraint to improve the Lipschitz continuity of the neural color field, making large changes of each Gaussian's color also match with large changes of its position in terms of a ratio. To this end, we additionally introduce novel self-learning strategies to improve the learning of the neural view-dependent color field without any external knowledge or priors, aiming for better generalization on unseen scenes. Our numerical and visual comparisons on widely used benchmarks justify our idea and show better ability of high fidelity geometry recovery in 3DGS. Project page: https://github.com/yangyixi21/Neural-Color-Fields-3DGS. |
|
| EAvatar: Expression-Aware Head Avatar Reconstruction with Generative Geometry Priors | 2026-11-02 | ShowHigh-fidelity head avatar reconstruction plays a crucial role in AR/VR, gaming, and multimedia content creation. Recent advances in 3D Gaussian Splatting (3DGS) have demonstrated effectiveness in modeling complex geometry with real-time rendering capability and are now widely used in high-fidelity head avatar reconstruction tasks. However, existing 3DGS-based methods still face significant challenges in capturing fine-grained facial expressions and preserving local texture continuity, especially in highly deformable regions. To mitigate these limitations, we propose a novel 3DGS-based framework termed EAvatar for head reconstruction that is both expression-aware and deformation-aware. Our method introduces a sparse expression control mechanism, where a small number of key Gaussians are used to influence the deformation of their neighboring Gaussians, enabling accurate modeling of local deformations and fine-scale texture transitions. Furthermore, we leverage high-quality 3D priors from pretrained generative models to provide a more reliable facial geometry, offering structural guidance that improves convergence stability and shape accuracy during training. Experimental results demonstrate that our method produces more accurate and visually coherent head reconstructions with improved expression controllability and detail fidelity. |
23 pages, 13 figures |
| Material-informed Gaussian Splatting for 3D World Reconstruction in a Digital Twin | 2026-11-02 | Show3D reconstruction for Digital Twins often relies on LiDAR-based methods, which provide accurate geometry but lack the semantics and textures naturally captured by cameras. Traditional LiDAR-camera fusion approaches require complex calibration and still struggle with certain materials like glass, which are visible in images but poorly represented in point clouds. We propose a camera-only pipeline that reconstructs scenes using 3D Gaussian Splatting from multi-view images, extracts semantic material masks via vision models, converts Gaussian representations to mesh surfaces with projected material labels, and assigns physics-based material properties for accurate sensor simulation in modern graphics engines and simulators. This approach combines photorealistic reconstruction with physics-based material assignment, providing sensor simulation fidelity comparable to LiDAR-camera fusion while eliminating hardware complexity and calibration requirements. We validate our camera-only method using an internal dataset from an instrumented test vehicle, leveraging LiDAR as ground truth for reflectivity validation alongside image similarity metrics. |
8 pag...8 pages, 5 figures. Accepted to IEEE Intelligent Vehicles Symposium (IV) 2026. Revised version (v4) presents camera-ready publication |
| GaussFlaw: Fine-Grained Perceptual Diagnosis for 3D Gaussian Splatting | 2026-11-02 | Show3D Gaussian Splatting (3DGS) enables photorealistic novel- view synthesis with real-time rendering, yet fine-grained di- agnosis of its rendering artifacts remains challenging. These artifacts are typically spatially localized and exhibit diverse degradation patterns, making them difficult to character- ize using global quality metrics and image-level assessment benchmarks, which provide no explicit indication of where they occur or what types of artifacts they represent. To ad- dress this limitation, we introduce GaussFlaw-80K, the first large-scale dataset for fine-grained perceptual diagnosis in 3DGS. Constructed through a model-assisted human-in-the- loop pipeline, GaussFlaw-80K contains 80,049 artifact re- gions across 14,289 rendered images, each annotated with a pixel-level mask and one of eight flaw labels spanning ge- ometric, blending, and photometric degradations. These an- notations turn 3DGS artifact diagnosis into an explicit and measurable task. Building on this dataset, we propose Ge- oCritic, a geometry-guided multi-task model that integrates RGB appearance with depth-derived cues to jointly localize artifacts and classify their flaw types. Beyond evaluation, we further demonstrate the downstream benefits of fine-grained diagnosis through three complementary targeted operations: Gaussian densification, attribute correction, and outlier prun- ing. Experiments show that GeoCritic achieves an AUC-J of 0.9507 and an Acc@1 of 80.08%, while diagnosis-guided re- finement improves all reported global and crop-level metrics across diverse datasets and 3DGS backbones. |
|
| OctreeSplat: Hierarchical Octree-Guided Feed-Forward Gaussian Splatting | 2026-11-02 | Show3D Gaussian Splatting (3DGS) has shown considerable potential for novel view synthesis (NVS) and 3D reconstruction. Feed-forward methods avoid costly per-scene optimization, but fixed image or spatial resolutions limit their ability to allocate primitives according to local complexity. We present OctreeSplat, a hierarchical octree-guided feed-forward Gaussian splatting. OctreeSplat extracts multi-view features and geometric priors from a frozen visual foundation model (VFM), constructs a hierarchical octree, and predicts densification scores. The resulting representation allocates Gaussian primitives at spatially varying densities according to local complexity. Stochastic solids volume rendering improves depth rendering, strengthens geometric supervision, and supports surface reconstruction. When camera parameters are available, an optional test-time optimization (TTO) module improves agreement between predicted and ground-truth parameters. Training updates only 76 million additional parameters and requires 35 GPU-hours. Experiments on Replica, TUM, Tanks and Temples, and DTU demonstrate competitive RGB view synthesis, depth synthesis, and surface reconstruction. OctreeSplat reduces the number of Gaussian primitives while maintaining high reconstruction quality. |
|
| SCARF: A Scene-Adaptive Depth-Guided G-3DGS Encoder Accelerator with Semantic Reuse and Fused Dataflow | 2026-11-02 | ShowDepth-guided, cost-volume-based Generalizable 3D Gaussian Splatting (G-3DGS) encoders reconstruct scenes as sets of Gaussian primitives directly from multi-view images without requiring per-scene optimization, thereby facilitating feed-forward novel-view synthesis on edge computing platforms. However, their shared encoder architecture presents two primary bottlenecks. First, the per-pixel depth search disregards semantic similarity within the 2D feature space, causing depth search operations to account for over 65% of memory traffic due to redundant exhaustive candidate scans. Second, the sequential pipeline of depth prediction followed by Gaussian generation neglects local 3D geometric continuity, resulting in more than 52% of generated Gaussians being redundant prior to instantiation. In this paper, we introduce SCARF, a dataflow-co-designed accelerator that aligns each bottleneck with a corresponding scene-level regularity. To mitigate the first bottleneck, Feature Similarity Depth Reuse reconceptualizes depth search as a problem of semantic locality: depth consistency is determined by affinity in feature space rather than physical address proximity. It transforms exhaustive per-pixel depth searches into anchor-centered local searches. To overcome the second bottleneck, Scene-Adaptive Early Sparsification disrupts the conventional serial depth-to-materialization boundary by probing each tile to infer local scene regularity. It then selectively activates only the minimally sufficient generation pathway, eliminating redundant Gaussian computations before primitive instantiation. On a 27.8 square millimeter 28 nm post-layout design, SCARF delivers a 2.94x geometric-mean end-to-end speedup with maximum PSNR degradation below 0.072%, 2.23 inferences per joule, and 34.5x GPU energy efficiency. |
17 pa...17 pages, 16 figures. Accepted by MICRO 2026; awarded all three MICRO 2026 Artifact Evaluation badges. Code repo: https://github.com/MAdrid1011/SCARF |
| RxGS: Receiver-Generalizable 3D Gaussian Splatting for Radio-Frequency Data Synthesis | 2026-11-02 | ShowRadio-frequency (RF) data synthesis predicts the received signal given transmitter and receiver positions, and is essential for wireless applications. Recent 3D Gaussian Splatting (3DGS)-based methods achieve efficient synthesis at any transmitter location within a surveyed scene, but require a separate model for each receiver. We therefore present RxGS, a shared 3DGS model that first learns receiver-independent Gaussian geometry and then freezes it while learning directional radiance conditioned on receiver position. Its global conditioning branch captures receiver-dependent effects shared across the scene, while a local branch models per-scatterer variations from the receiver's geometry and occlusion. The shared geometry also supports batched rendering through a multi-receiver rasterizer. Across RF datasets, RxGS predicts at receiver positions excluded from training within the surveyed scene while significantly reducing training, inference, and storage costs. Project page: https://kangyangg.com/projects/rxgs/. |
|
| FLEG: Feed-Forward Language Embedded Gaussian Splatting from Any Views via Compact Semantic Representation | 2026-11-01 | ShowWe present FLEG, a feed-forward network that reconstructs language-embedded 3D Gaussians from multi-view images. Previous feed-forward language-embedded Gaussian reconstruction methods are restricted to a fixed number of input views and typically attach a language-aligned semantic embedding to each Gaussian, resulting in impractical input settings and semantic redundancy. In contrast, we introduce a geometry-semantic dual-branch distillation framework that supports a variable number of uncalibrated and unposed multi-view images as input. We also propose a novel-view-based distillation strategy during training that mitigates overfitting to input views. In addition, we observe that semantic representations are significantly sparser than geometric ones, and per-Gaussian language embedding is unnecessary. To exploit this sparsity, we design a decoupled language embedding strategy that represents language information with a sparse set of semantic Gaussians, rather than attaching embeddings to every Gaussian. Compared with dense pixel-aligned per-Gaussian embedding schemes, our method uses only 5% of the language embeddings while maintaining comparable semantic fidelity, effectively reducing storage costs. Extensive experiments demonstrate that FLEG outperforms state-of-the-art feed-forward reconstruction and language-embedded Gaussian methods in both reconstruction quality and language-aligned semantic representation. |
Accep...Accepted by ECCV 2026. Project Page: https://fangzhou2000.github.io/projects/fleg |
| VLSplat: Vision-Language Guided Object-Centric 3D Gaussian Splatting via Scene Graph | 2026-11-01 | ShowLearning semantic representations for 3D Gaussian Splatting has recently emerged as a promising direction for 3D scene understanding. However, existing 2D-to-3D semantic lifting approaches rely primarily on 2D segmentation supervision and lack object-level priors, causing the learned representation to be driven by local mask evidence rather than holistic object structure. In this paper, we present VLSplat, a vision-language guided approach for object-centric semantic lifting in 3D Gaussian Splatting. Our method augments mask-based lifting with object-level semantic priors derived from vision-language models and refines Gaussian representations using an object-group scene graph. The refinement is performed through two language-guided modules: intra-group refinement suppresses semantic noise within object groups, while inter-group acquisition expands object support regions by incorporating structurally relevant candidates from neighboring groups. Extensive experiments on indoor scene datasets demonstrate that VLSplat improves the semantic consistency and structural completeness of object-level categories. Our code and models are available at https://vlsplat.github.io/. |
Accep...Accepted to NeurIPS 2026. Project page: https://vlsplat.github.io/ |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| Spade-SLAM: Sparse Correspondence Anchors for Dense Self-Supervised Visual SLAM | 2026-11-01 | ShowWe introduce Spade-SLAM, a self-supervised dense visual SLAM framework that substantially narrows the long-standing performance gap to supervised methods in a zero-shot setting. By demonstrating that high-fidelity tracking can be achieved without explicit ground-truth pose and depth labels, our approach offers a highly effective and label-efficient alternative to fully supervised pretraining. Our core contribution is a complementary sparse-to-dense training regime: While a warping-based photometric loss provides a dense training signal, we stabilize the optimization and suppress drift in ambiguous regions using a sparse correspondence loss derived from image feature matches. Extensive experiments demonstrate that Spade-SLAM significantly improves robustness and accuracy compared to existing self-supervised baselines, closely rivaling SLAM systems trained with explicit pose and depth supervision. The code will be made publicly available at https://github.com/JulianPSchmitt/Spade-SLAM. |
|
| Action-Consistent Decentralized Belief Space Planning with Inconsistent Beliefs and Limited Data Sharing: Framework and Simplification Algorithms with Formal Guarantees | 2026-10-31 | ShowIn multi-robot systems, ensuring safe and reliable decision making under uncertain conditions demands robust multi-robot belief space planning (MR-BSP) algorithms. While planning with multiple robots, each robot maintains a belief over the state of the environment and reasons how the belief would evolve in the future for different possible actions. However, existing MR-BSP works have a common assumption that the beliefs of different robots are same at planning time. Such an assumption is often unrealistic as it requires prohibitively extensive and frequent data sharing capabilities. In practice, robots may have limited communication capabilities, and consequently beliefs of the robots can be different. Crucially, when the robots have inconsistent beliefs, the existing approaches could result in lack of coordination between the robots and may lead to unsafe decisions. In this paper, we present decentralized MR-BSP algorithms, with performance guarantees, for tackling this crucial gap. Our algorithms leverage the notion of action preferences. The base algorithm VerifyAC guarantees a consistent joint action selection by the cooperative robots via a three-step verification. When the verification succeeds, VerifyAC finds a consistent joint action without triggering a communication; otherwise it triggers a communication. We design an extended algorithm R-VerifyAC for further reducing the number of communications, by relaxing the criteria of action consistency. Another extension R-VerifyAC-simp builds on verifying a partial set of observations and improves the computation time significantly. The theoretical performance guarantees are corroborated with simulation results in discrete setting. Furthermore, we formulate our approaches for continuous and high-dimensional state and observation spaces, and provide experimental results for active multi-robot visual SLAM with real robots. |
The n...The new version has been extended from the existing arxiv version of the paper in the following way: - Figures 2, 3, 10 are added; Section 6 - "Planning implications on MR-AC" is added; Section 8 - "Implementation and results" has been modified; Section 9 - "Conclusions" has been modified |
| CLEAR-IR: Clarity-Enhanced Active Reconstruction of Infrared Imagery | 2026-10-27 | ShowThis paper presents CLEAR-IR (Clarity Enhanced Active Reconstruction of Infrared Imagery), a novel framework for enabling robust robotic perception in dark environments using active infrared (IR) sensing. IR streams are less susceptible to noise than RGB in low light conditions but are dominated by structured light emitter patterns that hinder high level tasks such as object detection, tracking, and localisation. CLEAR-IR addresses this by reconstructing clean, RGB styled images from emitter populated IR inputs using a dual stream architecture inspired by DeepMAO, enabling standard RGB trained vision pipelines to operate without visible light. Evaluation across object detection, ArUco marker detection, and visual SLAM demonstrates that CLEAR-IR produces reliable perception in conditions ranging from well lit to completely dark, where passive RGB based enhancement methods fail entirely due to the physical absence of sensor signal. |
Accep...Accepted and presented in IEEE/RSJ IROS 2026. Project page: https://ladoszlabs.github.io/clearir/ |
| Bracketing Shows the Way: Adaptive Multi-Exposure Policy as Input for Robust Visual SLAM | 2026-10-23 | ShowRapid illumination transitions can disturb Visual SLAM (VSLAM) by saturating images before the Automatic Exposure (AE) policy adapts, removing the visual correspondences needed to maintain tracking. Our proposed methods mitigate this disturbance at two levels: B-Adapt acquires adaptive exposure brackets to cover the scene dynamic range while promoting photometric overlap, and BVSLAM directly uses the multi-exposure sequence as constraints in a factor graph state estimator without relying on High Dynamic Range (HDR) merging. We evaluate both methods against state-of-the-art approaches in nominal environments and during a repeated tunnel-exit transition at an emulated vehicle speed of 32 m s-1. In the tunnel, combining our B-Adapt with BVSLAM reduces the state estimation error by 78 % compared with AE images. These results suggest that bracketing efficiently improves visual pose estimation robustness to rapid illumination disturbances while retaining low error under nominal conditions. Our code is available at https://github.com/norlab-ulaval/bracketing-shows-the-way. |
9 pag...9 pages, 6 figures, submitted to 2027 IEEE International Conference on Robotics and Automation (ICRA 2027) |
| Evaluation of Front-End Feature Matching for Lunar Visual SLAM | 2026-10-22 | ShowLunar rovers will primarily rely on cameras for localization until navigation satellite services are deployed, and will continue to need them afterwards for relative navigation, hazard detection, and surface operations. Visual odometry (VO) and Simultaneous Localization and Mapping (SLAM) mostly estimate motion by matching features between frames. However, feature matchers are developed on urban, indoor and driving scenes, planetary studies compare them on orbital registration or within complete SLAM systems, and real lunar imagery with reference poses is scarce, so matching between rover frames on low-texture, harshly lit terrain, and its effect on VO, has not been compared. We benchmark eleven classical and learned detector-matcher combinations on nine synthetic, lunar-testbed, and terrestrial-analogue datasets, and a feature tracker, scoring pairwise pose accuracy against baseline and the trajectory error of the resulting VO. Under a common rotation-only metric, the SuperPoint pairings outperform SIFT with ratio-tested FLANN matching on both lunar testbeds and one analogue site, with smaller or unresolved differences on synthetic data. When rotation and translation direction are scored jointly, this SIFT pairing exceeds both SuperPoint pairings on four synthetic datasets, so synthetic results alone do not establish relative performance on real rover imagery. In stereo VO, the SuperPoint pairings are not consistently more accurate than SIFT, but only they keep tracking on all eight sequential datasets. Pairings with classical detectors depend on lighting while SuperPoint changes little: a low spotlight makes them more accurate than diffuse light on the same testbed views, consistent with more contrast and detected keypoints. SuperPoint with LightGlue is the fastest learned combination (16 ms per frame), and DINOv2 retrieval proposals still need geometric verification. |
16 pa...16 pages, 8 figures, 5 tables. Submitted to the 2027 IEEE Aerospace Conference |
| Hilti-Trimble-Oxford Dataset: 360 Visual-Inertial Benchmark with Floor Plan Priors for SLAM and Localization | 2026-10-21 | ShowAutomated progress monitoring on construction sites is an active area of research and development. Robot and human-carried mapping systems have been developed to build 3D maps of building and infrastructure projects. While LiDAR-based mapping systems achieve high accuracy, the cost of LiDAR can be prohibitive. Consumer-grade cameras with wide field of view ("360 cameras") combined with embedded inertial measurement units (IMUs) provide a cost-effective alternative. To support change detection and progress monitoring, highly accurate visual simultaneous localization and mapping (SLAM) and floor plan localization systems are required. In this paper, we present a high-quality dataset collected at an active construction site, which captures realistic challenges such as variable lighting conditions, moving workers, fast motions, and repetitive structures. The dataset offers thirty visual-inertial sequences recorded across ten floors over an eight-month period of the construction project. Ground truth trajectories are collected using a high quality LiDAR-inertial SLAM system rigidly attached to the 360 camera. Additionally, we report the results of an open research challenge evaluating the best visual SLAM and localization systems from around the world. The Challenge attracted substantially higher participation in SLAM, with 62 teams compared to 22 in floor plan localization, which suggests that this task receives less attention from researchers and motivates further exploration through proposed dataset and challenge, available at: https://hilti-trimble-challenge.com/dataset-2026. |
|
| HyVGGT-VO: Hybrid Dense Visual Odometry With Classical VO and Feed-Forward Model | 2026-10-16 | ShowDense visual odometry (VO), which provides pose estimation and dense 3D reconstruction, serves as the cornerstone for applications ranging from robotics to augmented reality. Recently, feed-forward models have demonstrated remarkable capabilities in dense mapping. However, when these models are used in dense visual SLAM systems, their keyframe-based processing often leads to sparse pose outputs, limiting continuous per-frame pose estimation. In contrast, traditional sparse methods provide high-frequency pose outputs but lack the capability for dense reconstruction. To address these limitations, we propose HyVGGT-VO, a novel framework that combines the computational efficiency of sparse VO with the dense reconstruction capabilities of feed-forward model. To the best of our knowledge, this is the first work to tightly couple a traditional VO frontend with VGGT. Specifically, we design an adaptive hybrid tracking frontend that dynamically switches between traditional optical flow and the VGGT model to ensure robustness. Furthermore, we introduce a hierarchical optimization framework that jointly refines VO poses and the scale of VGGT predictions to ensure global scale consistency. Under the odometry setting, compared with VGGT-SLAM 2.0, HyVGGT-VO reduces the average ATE by approximately 91% on EuRoC MAV and 20% on TUM RGB-D, and decreases the average relative translation and rotation errors by 46% and 92% on KITTI, respectively, while achieving 16.1 Hz pose output and 3.2 Hz dense-mapping updates with 6.5 GB of memory on an 8 GB laptop GPU.Our code is available at https://geneta2580.github.io/HyVGGT-VO.io. |
|
| Praxist: From Experimental Artifacts to Solution Lineages | 2026-10-15 | ShowAutonomous R&D agents now write, run, and improve executable artifacts under automated evaluation-but largely as laboratory instruments: shown on curated benchmarks, with gains that are hard to trace to a cause and costs well above what sustained engineering practice absorbs. The limitation is structural. Most systems treat each attempt as nearly self-contained, so logs, memories, and search trees record what happened without establishing which design element produced an improvement, whether its evidence survived validation, or how it recombines with others. Long campaigns therefore keep re-learning the same lessons. We introduce Praxist, a lineage-centered generational system that converts reproducible artifacts and evaluator outcomes into a typed evidence graph of findings, lane-structured frontiers, and agendas. Separating local artifact construction from cohort-level evidence synthesis lets later attempts inherit validated mechanisms, unresolved claims, and useful constraints, and leaves results attached to an inspectable lineage. On the standardized 75-task MLE-bench suite, the finalized official-grader results give Praxist 60 medals (80.0%), 49 of them gold, against 55 medals (73.3%) and 34 gold for a Claude Code baseline on Claude Opus 4.8-at a recorded model spend of $3054 versus $38370, roughly a twelfth of the cost. Five case studies-quantitative trading, LiDAR-inertial-visual SLAM, tokamak magnetic control, rocket landing, and job-shop rescheduling-carry the same process into open-ended engineering problems, improving on each task-native baseline in headline accuracy, survival, or resource cost, with the discovery path on record. Stronger artifacts at an order of magnitude less spend, each backed by an auditable lineage, are, to our knowledge, first brought together here: the operating profile production research requires, not the one a benchmark demonstration establishes. |
|
| Line-Augmented cuVSLAM: Robust Visual Odometry in Low-Texture Environments | 2026-10-14 | ShowAutonomous mobile robots in warehouses and production facilities operate in precisely the environments that defeat point-feature visual odometry: featureless aisles, uniform walls, repetitive racking and specular floors. Such environments are, however, rich in straight-edge structure: rack uprights, door frames, wall junctions and cabinet outlines present high-contrast linear edges stable across wide viewpoint changes. A line-feature extension to cuVSLAM, a CUDA-accelerated visual-SLAM system, is presented that recovers this signal. Lines are parameterised as Plücker coordinates for geometric transport and as the minimal 4-DOF orthonormal representation of Bartoli and Sturm for optimisation; a Cauchy-robust point-to-line residual with analytic Jacobians is folded directly into cuVSLAM's per-frame Gauss-Newton pose solve, and the same line state is jointly refined in the Schur-complement sliding-window bundle adjustment. An adaptive information weight causes lines to assume more of the estimation burden exactly as point features become scarce, and the extension adds residual blocks to existing solvers rather than introducing a second estimator, reducing exactly to the unmodified point pipeline when disabled. Evaluated in RGB-D mode under SE(3) Umeyama alignment with no scale correction, line features reduce mean absolute trajectory error by 38.5% over the 22 sequences of the TUM RGB-D Freiburg3 collection and by 30.1% over the 8 Replica room and office sequences, where median translation error falls by 38.9% and median rotation error by 66.4%, with zero tracking failures in any configuration tested. |
8 pages, 4 figures |
| RingNav: Task-Shaped Circular Memory for Visual Heading | 2026-10-13 | ShowTiny robots have limited computation and memory, yet visual navigation systems often maintain more state than a task requires. We ask whether persistent memory can instead be shaped by the task. We present \textbf{RingNav}, a (10.6)k--parameter visual heading estimator that constrains its memory to the circular manifold (S^1). RingNav learns only the turn between consecutive images, while heading is composed directly on the circle. On the TartanAir Visual SLAM Challenge, RingNav reduces mean absolute turn error by (27%) and accumulated heading error over |
9 pages, 5 figures |
| FoundationVIO: Enforcing Metric Scale in Feed-forward Visual-Inertial Odometry | 2026-10-10 | ShowFeed-forward geometric foundation models (GFMs) directly predict dense geometry and camera poses from batches of images. Emerging visual SLAM methods chain together consecutive GFM batches to form dense point cloud maps. A core challenge is how to address the ambiguous geometric scale of GFM predictions: independent predictions are locally consistent up-to-scale but not constrained by any direct physical measurements. Models trained on data with known scale improve the consistency of GFM reconstructions but again do not directly measure it. Consequently, trajectories and 3D reconstructions from GFM-enabled SLAM methods inherit this issue. Motivated by parallels with classical visual-inertial odometry, we introduce FoundationVIO which integrates inertial sensing with GFM predictions to recover metric scale reconstruction using a factor-graph based approach. We demonstrate odometry estimates competitive with classical state-of-the-art VIO methods and accurate dense reconstructions that outperform GFM-enabled SLAM methods. We also add a minimal GFM-based loop-closure mechanism to create a full Visual-Inertial SLAM system capable of creating globally accurate reconstructions and pose estimates. We evaluate our method on three datasets with well known sensor systems and demonstrate its compatibility on ubiquitous camera and IMU sensors. The method runs at approximately real-time on a consumer-grade laptop GPU. Code: https://github.com/ori-drs/FoundationVIO. |
|
| RIPE++: Reinforced Keypoint Learning from Positive Pairs Only | 2026-10-06 | ShowSparse keypoint extraction and matching underpin core tasks in geometric computer vision, including structure-from-motion, visual SLAM, augmented reality, and medical image registration. Learning robust local feature representations, however, typically requires accurate camera poses or depth supervision, which are often unavailable in real-world settings. Reinforcement learning (RL) has recently emerged as a promising alternative, requiring only the information if two images show the same scene or not. However, existing RL formulations such as RIPE rely on coarse binary rewards and carefully constructed negative training pairs, limiting training stability and descriptor discriminability. In this paper, we revisit RL-based keypoint learning and propose a reward that fully exploits the geometric consistency signal, deriving both reward and penalty from a single positive pair without contrasting against negatives. This richer signal provides sufficient supervisory contrast to learn discriminative detectors and descriptors from positive image pairs alone, enabling representation learning under extremely limited supervision. Furthermore, we show that the same RL objective can be extended to the matching stage by adapting LightGlue, raising AUC@5 on MegaDepth1500 from 56.58 to 59.65 and enabling weakly-supervised training of the full sparse matching pipeline from image pairs with partial visual overlap. We validate our approach on established benchmarks, demonstrating competitive results compared to fully-supervised methods. We further show that the method can be even trained on low texture medical video sequences, where camera poses are usually unavailable and standard SfM pipelines often fail. Code and data are available at https://github.com/fraunhoferhhi/RIPEpp . |
LIMIT...LIMIT@ECCV 2026 (Best Paper Award) |
| Dense Dynamic Scene Reconstruction and Camera Pose Estimation from Multi-View Videos | 2026-10-05 | ShowWe address the challenging problem of dense dynamic scene reconstruction and camera pose estimation from multiple freely moving cameras -- a setting that arises naturally when multiple observers capture a shared event. Prior approaches either handle only single-camera input or require rigidly mounted, pre-calibrated camera rigs, limiting their practical applicability. We propose a two-stage optimization framework that decouples the task into robust camera tracking and dense depth refinement. In the first stage, we extend single-camera visual SLAM to the multi-camera setting by constructing a spatiotemporal connection graph that exploits both intra-camera temporal continuity and inter-camera spatial overlap, enabling consistent scale and robust tracking. To ensure robustness under limited overlap, we introduce a wide-baseline initialization strategy using feed-forward reconstruction models. In the second stage, we refine depth and camera poses by optimizing dense inter- and intra-camera consistency using wide-baseline optical flow. Additionally, we introduce MultiCamRobolab, a new real-world dataset with ground-truth poses from a motion capture system. Finally, we demonstrate that our method significantly outperforms state-of-the-art feed-forward models on both synthetic and real-world benchmarks, while requiring less memory. |
fix a...fix author name errors |
| Human-in-the-Loop Neuro-Symbolic Drift Anticipation for Reliable Visual SLAM | 2026-10-05 | ShowThis paper introduces Hybrid DeepSEE (HDS), a Human-in-the-Loop (HITL) neuro-symbolic framework for proactive drift anticipation in Visual SLAM (V-SLAM). While data-driven models offer predictive power, their "black-box" nature often yields physically inconsistent outputs in out-of-distribution (OOD) environments. To address this, HDS integrates neural drift risk estimation with symbolic constraint reasoning. By utilizing a Large Language Model (LLM) as a reasoning bridge, the framework translates qualitative human context into interpretable symbolic constraints. Building upon this architecture, we propose a superior drift anticipation framework that ensures enhanced reliability and consistency in Visual SLAM |
IFAC ...IFAC conference paper; 5 figures |
| Does Dynamic-Point Filtering Help When Texture Is Scarce? A Controlled Study of ORB-SLAM2 Front-Ends in Synthetic Indoor Scenes | 2026-10-02 | ShowDynamic-point filters are routinely added to feature-based visual SLAM, and several recent systems argue that removing dynamic features can leave too few static features in low-texture regions. So far, these systems have been evaluated only on texture-rich benchmark sequences. We present a controlled study that isolates this interaction. We render synthetic indoor sequences in which surface texture (four levels, quantified by FAST-corner density and image-gradient entropy) and scene dynamics (three levels) are varied factorially along identical camera trajectories, with stereo, RGB-D, ground-truth poses and dynamic masks. On this grid we compare ORB-SLAM2 without filtering, with an optical-flow and epipolar-residual filter (FLOW), and with a multi-view depth-consistency filter (GEOM), and report trajectory error, tracking completeness and surviving static features over five runs. Because the masks give per-keypoint ground truth, we also measure each filter's dynamic-point precision and recall and its static-feature false-removal rate, so that mechanistic explanations can be tested directly. We do not propose a new filter. On 720 runs over 24 sequences, filtering helped mainly in the most dynamic cells; the benefit did not decline monotonically with texture, but at the lowest level filtering reduced tracking completeness, and ORB-SLAM2 never initialised in static L3 scenes. Contrary to our hypothesis, GEOM discarded more static keypoints than FLOW (median FRR 6.8% vs. 1.5% for RGB-D, 19.6% vs. 1.5% for stereo); its RGB-D advantage tracked dynamic-point recall and vanished in stereo mode. Data and code are available at https://github.com/felixxxue/texture-dynamics-slam. |
8 pag...8 pages, 4 figures, submitted to IROS 2026. Open-source code and dataset available |
| BIM Informed Visual SLAM for Construction Environments | 2026-09-30 | ShowMonitoring building construction sites requires comparing the as-planned design with the as-built state, which can be estimated in real time using Simultaneous Localization and Mapping (SLAM) techniques. However, visual SLAM is prone to trajectory drift in construction environments, producing maps that are geometrically inaccurate with the actual environment. To address this limitation, we augment an existing RGB-D SLAM system with structural priors derived from the Building Information Model (BIM). The system associates detected walls with their BIM counterparts and includes these correspondences as geometric constraints in the back-end optimization, reducing drift and enhancing global consistency. The proposed method operates in real time and is validated on multiple real construction sites, achieving an average trajectory error reduction of 25.23% and a 7.14% improvement in map accuracy over state-of-the-art baselines. Robustness analyses further demonstrate resilience to incomplete BIM data and geometric discrepancies between as-planned models and the as-built environment. |
8 pag...8 pages, 7 tables, 4 figures |
| MVP-SLAM: Multi-Camera Visual-Inertial Floorplan-Prior SLAM | 2026-09-30 | ShowIndoor building construction sites are demanding environments for visual SLAM, where variable lighting and repetitive, low-textured structures make the system drift over long trajectories, though structural elements such as walls remain distinguishable despite these conditions. These buildings are constructed according to their as-planned floor plans, available from the design phase, and although the actual as-built site can differ from this design, floor plans still provide a metric reference, both to localize the system in the building and to correct drift. Existing methods often use the floor plan to correct an already-built trajectory offline, and those that instead correct it online typically rely on depth sensors. We instead present MVP-SLAM, an online visual-inertial SLAM on two opposite-facing fisheye cameras that corrects drift from cameras alone by matching walls detected in its map to the floor plan, through a drift-aware policy. A multi-stage integration then turns each matched pair incrementally into a persistent correction, so the trajectory stays corrected and localized within the floor plan as it is built. MVP-SLAM was validated on the multi-floor construction sites of the Hilti-Trimble SLAM Challenge 2026, ranking 2nd of 22 teams in the Localization task (0.29 m mean RMSE) and 5th of 62 teams in the SLAM task (0.24 m), the top-ranked one in both tasks among those that operate online, integrate the floor plan, and localize within it. |
9 pag...9 pages, 4 figures. Submitted to IEEE Robotics and Automation Letters (RA-L) |
| ScaRF-SLAM: Scale-Consistent Reconstruction with Feed-Forward Models and Classical Visual SLAM | 2026-09-25 | ShowRecent works have explored unifying SLAM with geometric foundation models (GFMs). However, directly using GFM predictions for tracking is highly sensitive to model capability and uncertainty, as geometric inaccuracies in the predictions can adversely affect pose estimation. To address this limitation, we propose a decoupled framework that integrates classical feature-based SLAM with GFMs, which achieves higher quality and more consistent dense reconstruction. In brief, we use classical visual SLAM for robust low-latency tracking and use GFMs exclusively for mapping. By anchoring mapping to poses produced by the SLAM module and optimizing across depth scales, the proposed design avoids propagating inaccuracies from GFM predictions into pose estimation while imposing geometric constraints on the reconstruction. The system builds submaps from multiple posed keyframes and enforces scale consistency via lightweight frame and submap scale optimization. It also performs projection-based point cloud fusion within each submap, and updates submaps online to reflect trajectory updates from the feature-based SLAM. To evaluate tracking and reconstruction of our method, we introduce a loop-rich, building-scale indoor dataset with accurate sensor trajectories and LiDAR ground-truth. Experiments show that our approach achieves superior trajectory accuracy while improving reconstruction precision by 10%-20% over existing methods, with about 2 cm reconstruction error per 10 m chunk on building-scale dataset. On large-scale outdoor datasets, it attains 10 cm error per 30 m chunk (w.r.t LiDAR ground-truth models). Code and dataset: https://github.com/ori-drs/ScaRF-SLAM |
Accep...Accepted to IEEE Robotics and Automation Letters (RA-L) 2026 |
| Know-Your-Scene (KYS)-SLAM: Hierarchical Semantic-Motion Priors for Feature Matching in Stereo Visual SLAM | 2026-09-23 | ShowStereo visual SLAM systems built on local descriptors suffer from semantic ambiguity, instance-level confusion, and independently moving objects, each corrupting data association and accumulating as trajectory drift. Prevailing semantic and dynamic SLAM methods address this through binary feature rejection, sacrificing correspondence density for outlier suppression. We contend that contextual implausibility is better expressed as a graded quantity than an exclusion criterion. We present Know-Your-Scene (KYS)-SLAM, a modular extension of ORB-SLAM3 that supplants feature rejection with continuous correspondence modulation. The contribution is the reframing of contextual evidence as correspondence cost, applied within feature matching and leaving the geometric backend unmodified. Each keypoint is augmented with semantic, panoptic, and motion priors fused through a hierarchical compatibility formulation, in which semantic class and instance identity enforce structural plausibility while a zero-shot motion score down-weights features on independently moving objects. That score comes from a training-free module fitting a depth-aware ego-motion model to background optical flow and classifying panoptic segments via self-calibrating, coverage-aware thresholds, so only segments with sufficient motion evidence are penalized and static structure is left unpenalized. Penalizing correspondences rather than discarding them preserves the geometric support bundle adjustment depends on. Under one fixed configuration, no coefficient retuned per sequence or dataset, KYS-SLAM reduces per-sequence ATE RMSE by 17.4% on outdoor KITTI and 27.7% on indoor EuRoC across 21 stereo sequences with no regressions, and by 6.6% on dynamic subsets of KITTI Tracking and 17.8%, up to 31.2%, on Virtual KITTI 2 -- cross-domain transfer across outdoor driving, indoor flight, and synthetic imagery under one set of constants. |
16 pa...16 pages, 9 figures, 12 tables; includes supplementary material |
| Why does Deep Learning Improve Visual SLAM? | 2026-09-21 | ShowVisual SLAM is a well-established technology utilized in a wide range of real-world applications. However, its performance still degrades under challenging visual conditions, such as low texture, severe motion blur, and poor illumination. Systems based on deep learning outperform classical geometry-based ones and achieve state-of-the-art results by combining learned 2D data association and uncertainty with differentiable geometric optimization in recurrent architectures. Still, it remains unclear exactly which components are fundamentally responsible for this success. In this paper, we ask: Is the superior performance of deep learning-based systems driven primarily by learned 2D data association, the combination of learned 2D data association and uncertainty, or the recurrent architecture itself? We investigate this question empirically by conducting a controlled study. Our findings reveal that the success of DL-based V-SLAM systems hinges on learned 2D data association and uncertainty rather than their recurrent architecture, underscoring the necessity of learning-based paradigms for the design of these components. |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| SVIn+: Multi-Modal Framework for Robust Underwater State Estimation using 3D Sonar, Visual-Inertial, Water-Pressure, and DVL | 2026-10-27 | ShowAs expansion and scalability in underwater oper- ations becomes more important and challenging, autonomous underwater vehicles (AUVs) become more relevant and the necessity for reliable autonomy more critical. A fundamental challenge in any autonomous underwater operation, from control to safe navigation and mapping, is state estimation; a notorious hard problem to address, especially in low visibility conditions. In this paper, we tackle this problem, by presenting SVIn+, a multi-sensor framework for robust state estimation of AUVs equipped with an IMU, cameras, a water-pressure sensor, a Doppler Velocity Log (DVL), and a compact 3D sonar. This framework extends an established open-source state-of-the-art visual-inertial SLAM, SVIn, by integrating water-pressure and DVL measurements into a tightly-coupled nonlinear optimiza- tion formulation resulting in improved estimates. To leverage the structural information of recently available 3D acoustic sonars while maintaining computational efficiency, we introduce a hierarchical architecture that incorporates estimates from a decoupled front-end via KISS-ICP, treated as relative pose factors within the factor graph. The proposed framework is validated through field deployments in a shipwreck and coral reef, demonstrating drift mitigation and global consistency. To support further research in underwater multi-modal fusion, the code and the datasets can be found on project website https://cmburgul.github.io/projects/svin+/ https://dartmouthrobotics.github.io/svin-plus |
IROS 2026 |
| GOFusion: Real-Time Gaussian-Occupancy Submap Fusion for Fisher-Informed Autonomous Robot Navigation | 2026-10-18 | ShowIn this letter, we present GOFusion, a real-time visual mapping system for autonomous robot navigation that anchors Gaussian primitives to local occupancy maps, providing both spatial awareness for path planning and high-fidelity appearance that enables information-driven robot navigation. GOFusion builds 3D Gaussian maps from pose estimates and local occupancy maps obtained via visual-inertial SLAM and enables pose corrections to maintain both geometric and visual consistency. Furthermore, it restricts the 3D Gaussian Splatting optimization to active submaps to bound GPU memory usage. Building on the appearance-rich Gaussian map, we introduce a rasterizer-integrated Fisher approximation that quantifies per-primitive sensitivity online, serving as an information-driven heuristic for Gaussian pruning and next-best-view planning. We evaluate GOFusion on public datasets and in real-world drone flights. It not only outperforms several baselines in mapping quality, but also achieves real-time, high-fidelity and uncertainty-aware mapping at scale. Finally, we demonstrate fully autonomous drone exploration for Gaussian map refinement using the proposed Fisher-informed heuristic. Code is available at https://github.com/Bennnoo/GOFusion. We provide a supplementary video demonstration at https://youtu.be/pLQKbxQvb6I. |
8 pag...8 pages, 7 figures, 4 tables. Paper under review |
| PanoAir: A Panoramic Visual-Inertial SLAM with Cross-Time Real-World UAV Dataset | 2026-10-16 | ShowAccurate pose estimation is fundamental for unmanned aerial vehicle (UAV) applications, where Visual-Inertial SLAM (VI-SLAM) provides a cost-effective solution for localization and mapping. However, existing VI-SLAM methods mainly rely on sensors with limited fields of view (FoV), which can lead to drift and even failure in complex UAV scenarios. Although panoramic cameras provide omnidirectional perception to improve robustness, panoramic VI-SLAM and corresponding real-world datasets for UAVs remain underexplored. To address this limitation, we first construct a real-world panoramic visual-inertial dataset covering diverse flight conditions, including varying illumination, altitudes, trajectory lengths, and motion dynamics. To achieve accurate and robust pose estimation under such challenging UAV scenarios, we propose a panoramic VI-SLAM framework that exploits the omnidirectional FoV via the proposed panoramic feature extraction and panoramic loop closure, enhancing feature constraints and ensuring global consistency. Extensive experiments on both the proposed dataset and public benchmarks demonstrate that our method achieves superior accuracy, robustness, and consistency, completing all 17 sequences without tracking failure and reducing the ATE by up to 47.1% compared with the best-performing baseline. Moreover, deployment on an embedded platform validates its practical applicability, achieving comparable computational efficiency to PC implementations. The source code and dataset are publicly available at https://github.com/YYWumm/PanoAir |
|
| RootSLAM: Square Root Bundle Adjustment for Visual-Inertial SLAM | 2026-10-10 | ShowEfficient perception is a prerequisite for the small, low-power devices targeted by robotics, mixed reality, autonomous driving, and drone applications, all of which rely on Simultaneous Localization and Mapping (SLAM) to estimate their motion within their surroundings. Learning-based SLAM systems achieve state-of-the-art accuracy but are demanding in compute, whereas classical geometry-based methods retain a strong efficiency advantage with little to no loss in accuracy. In this work we present RootSLAM, a fully-featured visual-inertial SLAM system for multi-camera setups built on Basalt. RootSLAM achieves accuracy competitive with the state of the art while running up to 20 times faster, facilitating the deployment of perception systems in resource constrained environments. We adapt the square-root bundle adjustment formulation from structure from motion into the multi-camera visual-inertial SLAM setting through a novel reformulation that leverages the system's pose-graph, letting the global optimizer run in single-precision floating point, thus doubling its performance. We provide a thorough benchmark against the state of the art together with an ablation study of our contributions. Our new implementation, including its map management and loop detection module, pose-graph optimizer, and global mapping components, will be open-sourced. |
|
| FoundationVIO: Enforcing Metric Scale in Feed-forward Visual-Inertial Odometry | 2026-10-10 | ShowFeed-forward geometric foundation models (GFMs) directly predict dense geometry and camera poses from batches of images. Emerging visual SLAM methods chain together consecutive GFM batches to form dense point cloud maps. A core challenge is how to address the ambiguous geometric scale of GFM predictions: independent predictions are locally consistent up-to-scale but not constrained by any direct physical measurements. Models trained on data with known scale improve the consistency of GFM reconstructions but again do not directly measure it. Consequently, trajectories and 3D reconstructions from GFM-enabled SLAM methods inherit this issue. Motivated by parallels with classical visual-inertial odometry, we introduce FoundationVIO which integrates inertial sensing with GFM predictions to recover metric scale reconstruction using a factor-graph based approach. We demonstrate odometry estimates competitive with classical state-of-the-art VIO methods and accurate dense reconstructions that outperform GFM-enabled SLAM methods. We also add a minimal GFM-based loop-closure mechanism to create a full Visual-Inertial SLAM system capable of creating globally accurate reconstructions and pose estimates. We evaluate our method on three datasets with well known sensor systems and demonstrate its compatibility on ubiquitous camera and IMU sensors. The method runs at approximately real-time on a consumer-grade laptop GPU. Code: https://github.com/ori-drs/FoundationVIO. |
|
| MVP-SLAM: Multi-Camera Visual-Inertial Floorplan-Prior SLAM | 2026-09-30 | ShowIndoor building construction sites are demanding environments for visual SLAM, where variable lighting and repetitive, low-textured structures make the system drift over long trajectories, though structural elements such as walls remain distinguishable despite these conditions. These buildings are constructed according to their as-planned floor plans, available from the design phase, and although the actual as-built site can differ from this design, floor plans still provide a metric reference, both to localize the system in the building and to correct drift. Existing methods often use the floor plan to correct an already-built trajectory offline, and those that instead correct it online typically rely on depth sensors. We instead present MVP-SLAM, an online visual-inertial SLAM on two opposite-facing fisheye cameras that corrects drift from cameras alone by matching walls detected in its map to the floor plan, through a drift-aware policy. A multi-stage integration then turns each matched pair incrementally into a persistent correction, so the trajectory stays corrected and localized within the floor plan as it is built. MVP-SLAM was validated on the multi-floor construction sites of the Hilti-Trimble SLAM Challenge 2026, ranking 2nd of 22 teams in the Localization task (0.29 m mean RMSE) and 5th of 62 teams in the SLAM task (0.24 m), the top-ranked one in both tasks among those that operate online, integrate the floor plan, and localize within it. |
9 pag...9 pages, 4 figures. Submitted to IEEE Robotics and Automation Letters (RA-L) |
| Towards Effective Visual-Inertial SLAM with Passive-Only Sensors for Low-Cost Autonomous Underwater Vehicles | 2026-09-17 | ShowImprovements to Visual-Inertial Simultaneous Localization and Mapping (VI-SLAM) for low-cost autonomous underwater vehicles (AUVs) are critical for transitioning advanced marine robotics from specialized labs to broader research and hobbyist applications. While high-end AUVs typically rely on expensive sensor suites - such as Doppler Velocity Logs (DVLs) and Ultra-Short Baseline (USBL) systems - this work demonstrates that robust, high-quality navigation is achievable using a sub-$10, 000(USD) platform equipped only with inexpensive consumer-grade sensors. By leveraging a similarly priced, open-source AUV, we evaluate the performance of stereo cameras, Micro-electromechanical System (MEMS)-based IMUs, and depth sensors in a fully unconstrained 6-degree-of-freedom (6-DOF) underwater environment. We analyze the efficacy of off-the-shelf SLAM packages and propose optimizations for sensor fusion to mitigate the visual and physical challenges of untethered underwater operation. Our results prove that a usable SLAM solution can be accessible to the masses, providing a benchmark for expectations in demanding, real-time maritime missions without the financial barrier of industrial-grade hardware. |
|
| AERO-VIS: Asynchronous Event-based Real-time Onboard Visual-Inertial SLAM | 2026-09-15 | ShowThe robustness of event cameras to high dynamic range and motion blur holds the potential to improve visual odometry systems in challenging environments. Although their high temporal resolution does not require synchronous processing, most event-based odometry methods still run at fixed rates, which simplifies system design but restricts latency and throughput. In this work, we present AERO-VIS, a stereo event-inertial SLAM system with an integrated, data-driven, robust, and performance-optimized keypoint detector. By decoupling event preprocessing from state estimation, our architecture operates asynchronously at the maximum rate supported by available hardware, dynamically adapting to downstream runtime demands and ensuring low-latency, real-time performance. When deploying AERO-VIS on a UAV, we achieve unprecedented accuracy in onboard event-based SLAM. These unique characteristics enable us to present the first purely event-based inertial SLAM system that demonstrates closed-loop UAV control and large-scale state estimation while relying solely on onboard compute. Videos of the experiments, source code, and additional results are available at https://ethz-mrl.github.io/AERO-VIS/. |
8 pages, 5 figures |
| Jetson-ORB-SLAM3: Accuracy-Preserving GPU Implementation for Edge Computing Devices | 2026-08-18 | ShowVisual-inertial SLAM on low-power edge platforms is constrained by the cost of dense feature extraction and loop closure. Prior GPU ports of ORB-SLAM trade accuracy for speed by approximating the ORB detector, altering the feature set and therefore the estimated trajectory. We present an accuracy-preserving GPU implementation of ORB-SLAM3 for the NVIDIA Jetson Orin Nano, whose GPU ORB front end reproduces the reference CPU detector algorithmically to 94.7% exact keypoint agreement and 99.9% descriptor bit agreement. This work also makes CNN-based loop closure edge-viable through native TensorRT. The visual front end (feature extraction) is offloaded to the GPU while the mapping and optimization back end is kept on the CPU, matching each computation to the hardware it suits. The accuracy is verified by comparing four configurations: the GPU pipeline and the unmodified CPU reference, each run on both the Jetson Orin Nano and a desktop. On EuRoC dataset, all four agree to within 0.10cm in mean absolute trajectory error (SE(3)), so neither the GPU port nor the change of hardware shifts the estimated trajectory. The GPU-versus-CPU comparison is reproducible on TUM-VI and KITTI datasets, so the acceleration is accuracy-preserving rather than approximate. The proposed implementation is competitive with published ORB-SLAM3 on EuRoC, attains sub-centimeter accuracy on five of the six TUM-VI room sequences, and reaches sub-1% relative translation error on nine of eleven KITTI sequences. For loop closure, the generic ONNX-Runtime CUDA/TensorRT execution providers are unusable with our CosPlace ResNet-50 on the embedded platform, whereas a native libnvinfer FP16 engine reduces per-query inference to 2.2ms, a 180x speedup. Learned place recognition therefore runs concurrently with tracking on a 7W device. In monocular-inertial mode the system sustains 32FPS mean over the eleven EuRoC sequences. |
|
| Mapping Pamir: Multi-Session Visual-Inertial SLAM and 3D Reconstruction of an Underwater Shipwreck | 2026-07-12 | ShowThis paper presents a framework for multi-session mapping of underwater environments utilizing an affordable action camera. The Visual-Inertial data are augmented by water depth recordings from a dive computer. SVIn2, an open-source VI-SLAM framework, is utilized to generate a trajectory and a sparse reconstruction for each session. Utilizing the keyframes extracted from SVIn2 and the estimated camera poses, a Structure-from-Motion (SfM) framework, COLMAP, is employed for global optimization and to produce a dense reconstruction of the target environment. The presence of calibration targets at fixed locations, when available, is used to estimate the coordinate transformation between different data collection sessions, thus transforming the different sessions into the same coordinate frame. The proposed pipeline is employed for the mapping of a shipwreck off the coast of Barbados. For the first time, both the exterior and the accessible interior parts of the wreck were mapped in two sessions, while a third session employed two cameras with different fields of view. |
8 pag...8 pages, 12 figures. Accepted to ICRA 2026, Vienna, Austria |
| FastTrack: GPU-Accelerated Tracking for Visual SLAM | 2026-07-08 | ShowThe tracking module of a visual-inertial SLAM system processes incoming image frames and IMU data to estimate the position of the frame in relation to the map. It is important for the tracking to complete in a timely manner for each frame to avoid poor localization or tracking loss. We therefore present a new approach which leverages GPU computing power to accelerate time-consuming components of tracking in order to improve its performance. These components include stereo feature matching and local map tracking. We implement our design inside the ORB-SLAM3 tracking process using CUDA. Our evaluation demonstrates an overall improvement in tracking performance of up to 2.8x on a desktop and Jetson Xavier NX board in stereo-inertial mode, using the well-known SLAM datasets EuRoC and TUM-VI. |
Publi...Published at IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2025) |
| PLED-VINS: A Point-Line Event-Based Visual Inertial SLAM for Dynamic Environments | 2026-07-08 | ShowDynamic environments remain a fundamental challenge for visual SLAM, where unreliable observations from moving objects and rapid motion degrade state estimation accuracy. Although event cameras preserve fine-grained spatio-temporal information, most existing event-based SLAM frameworks still assume static scenes and lack approaches to estimate the reliability of features. To this end, we propose PLED-VINS, a monocular event camera-based visual-inertial SLAM framework that enables robust state estimation in dynamic environments. We propose an entropy-recency score map to characterize the temporal reliability of both point and line features based on event temporal statistics. Concurrently, geometric reliability is estimated via a unified point-line robust bundle adjustment. Building upon these, we design an adaptive weighting strategy that fuses temporal and geometric reliability, including motion-conditioned reliability modeling for line features, to suppress unreliable observations. Experimental results demonstrate that PLED-VINS improves state estimation on the evaluated dynamic sequences with moving objects. |
8 pag...8 pages, 9 figures. Accepted to IROS 2026 |
| APVI-SLAM: Real-Time Acoustic-Pressure-Visual-Inertial Localization and Photorealistic Mapping System in Complex Underwater Environment | 2026-07-07 | ShowExtreme subsea environments often cause severe feature de-gradation and estimator divergence in underwater visual-inertial SLAM. Although sensors like Doppler Velocity Logs (DVL) and pressure gauges provide auxiliary constraints, robust multi-sensor fusion during intermittent visual failure remains challenging. To address this, we present APVI-SLAM, a real-time multi-sensor fusion SLAM system that achieves both accurate underwater localization and photorealistic mapping. Our approach introduces a reliability-aware localization framework that dynamically reweights sensor estimators and employs a sliding-window freezing strategy to recover from tracking failures, substantially enhancing system robustness. Furthermore, for high-fidelity scenes reconstruction, we propose an efficient quadtree-guided mapping module that facilitates incremental water-medium modeling and 3D Gaussian optimization. Recognizing the lack of benchmark for underwater mapping evaluation, we also contribute a coral reef surveying dataset with synchronized multi-modality data. Extensive experiments on public and our proposed benchmarks demonstrate that APVI-SLAM achieves state-of-the-art localization and reconstruction quality at real-time speeds. |
|
| DIVO: Continuous-time DVL-Inertial-Visual Odometry for Unmanned Underwater Vehicles | 2026-07-06 | ShowThis paper presents a novel acoustic-visual-inertial odometry solution leveraging a continuous-time trajectory estimation framework for unmanned underwater vehicles. Underwater environments present unique challenges for visual localization and mapping, such as light attenuation, illumination variance, and the presence of particulate matter. This motivates the use of additional sensing modalities and a visual tracking pipeline that is robust to diverse subsea conditions. The proposed system is the first continuous-time trajectory estimation framework based on Gaussian processes to fuse asynchronous measurements from a Doppler velocity log, a stereo camera, and an inertial measurement unit. Additionally, a novel visual frontend is proposed, incorporating learning-based feature extraction and matching that is robust to the specific challenges that subsea environments present. The proposed framework enables seamless integration of additional sensor modalities in continuous-time and is adaptable to different environments without reconfiguration. The proposed system is extensively tested on real-world underwater inspection datasets, where it outperforms state-of-the-art visual-inertial and acoustic-visual-inertial SLAM algorithms in accuracy, robustness, and trajectory coverage. Notably, the proposed system outperforms the state-of-the-art despite only forming short-term visual data associations. |
|
| DL-VINS-Factory: A Modular Framework for Learned Visual Front-Ends in Visual-Inertial SLAM | 2026-07-02 | ShowDeep-learning features excel in visual matching, yet their practical value in tightly coupled visual-inertial SLAM (VI-SLAM) remains insufficiently characterized. We present DL-VINS-Factory, a unified framework that integrates learned feature extractors (ALIKED, RaCo, SuperPoint, XFeat) with either Lucas--Kanade (LK) optical-flow tracking or LightGlue (LG) descriptor matching. All front-ends share a sliding-window Ceres back-end, with optional AnyLoc DINOv2-VLAD loop closure, and 4-DoF pose-graph optimization. We benchmark the system across the four datasets covering indoor, unstructured outdoor, aggressive-motion, and visually degraded conditions. Results show that learned front-ends are viable for real-time embedded VI-SLAM, but are not universally superior to classical tracking. Relative to the corresponding GFTT+LK baseline, ALIKED+LG reduces EuRoC ATE by |
|
| GeoFlow-SLAM++: A Robust Multi-Camera Visual-Inertial SLAM System with Relocalization | 2026-06-24 | ShowMonocular and RGB-D visual-inertial SLAM systems remain susceptible to limited field of view, sensor-specific failure modes, and unreliable cross-session relocalization. To address these issues, we present GeoFlow-SLAM++, a tightly coupled multi-camera visual-inertial SLAM system that extends GeoFlow-SLAM from a single RGB-D sensor to a calibrated multi-camera rig with a unified body-centric formulation. Within this multi-camera framework, GeoFlow-SLAM++ supports two interchangeable visual front-ends: a conventional ORB front-end and a neural network feature (NN-Feature) front-end built on SuperPoint and LightGlue. The system unifies tracking, mapping, and relocalization on a shared body state, and combines multi-camera reprojection constraints, IMU pre-integration, cross-view place recognition, and dual-stream optical flow/NN-Feature tracking for robust localization. As an optional extension, the system can further incorporate cross-view-consistent pseudo-depth predictions from RGB images as auxiliary geometric constraints. We evaluate GeoFlow-SLAM++ on EuRoC, OpenLORIS, TUM, Hilti, and a self-collected handheld multi-camera dataset. Results show that the NN-Feature front-end improves robustness in appearance-challenging scenarios, the multi-camera formulation achieves competitive localization accuracy on Hilti, and the unified cross-view relocalization design reaches LiDAR-comparable performance on the handheld dataset. |
10 pages |
| Sensor Configuration Matters: A Systematic Evaluation of Multimodal SLAM on Quadruped Robots | 2026-06-17 | ShowAutonomous navigation of quadrupedal robots in diverse environments fundamentally relies on resilient Simultaneous Localization and Mapping (SLAM). While visual-inertial SLAM has matured across wheeled, handheld, and aerial platforms, a critical evaluation gap remains regarding how hardware-level sensor configurations affect performance under the aggressive dynamics of legged locomotion. Quadrupeds introduce distinct embodiment-induced sensory challenges, including foot-impact shocks, high-frequency mechanical vibrations, and rapid angular rotations, which degrade standard perception pipelines. To address this gap, we present a systematic evaluation of state-of-the-art visual, visual-inertial, and LiDAR-visual-inertial SLAM methods using the GrandTour dataset recorded on an ANYmal D quadruped. We isolate and quantify the impacts of camera modalities, shutter techniques, and inertial sensor tiers, analyzing their trade-offs across localization accuracy, algorithmic robustness, and computational resource utilization. Our empirical findings demonstrate that hardware selection has substantial influence on system resilience: stereo configurations consistently outperform monocular and RGB-D modalities, global shutter cameras significantly mitigate motion-induced tracking failures compared to rolling shutter cameras, and, crucially, standard inertial integration can degrade the performance of primarily vision-based frameworks under harsh legged locomotion. These insights additionally offer concrete design guidelines for tailoring custom sensor payloads to achieve dependable perception on agile legged systems. |
|
| Omni-LIVO: Robust RGB-Colored Multi-Camera Visual-Inertial-LiDAR Odometry via Photometric Migration and ESIKF Fusion | 2026-03-29 | ShowWide field-of-view (FoV) LiDAR sensors provide dense geometry across large environments, but existing LiDAR-inertial-visual odometry (LIVO) systems generally rely on a single camera, limiting their ability to fully exploit LiDAR-derived depth for photometric alignment and scene colorization. We present Omni-LIVO, a tightly coupled multi-camera LIVO system that leverages multi-view observations to comprehensively utilize LiDAR geometric information across extended spatial regions. Omni-LIVO introduces a Cross-View direct alignment strategy that maintains photometric consistency across non-overlapping views, and extends the Error-State Iterated Kalman Filter (ESIKF) with multi-view updates and adaptive covariance. The system is evaluated on public benchmarks and our custom dataset, showing improved accuracy and robustness over state-of-the-art LIVO, LIO, and visual-inertial SLAM baselines. Code and dataset will be released upon publication. |
Accep...Accepted by IEEE Robotics and Automation Letters (RA-L). Early Access version available. This version supersedes all previous versions and is the official accepted manuscript for citation |
| Edge-Assisted Multi-Robot Visual-Inertial SLAM with Efficient Communication | 2026-03-11 | ShowThe integration of cloud computing and edge computing is an effective way to achieve global consistent and real-time multi-robot Simultaneous Localization and Mapping (SLAM). Cloud computing effectively solves the problem of limited computing, communication and storage capacity of terminal equipment. However, limited bandwidth and extremely long communication links between terminal devices and the cloud result in serious performance degradation of multi-robot SLAM systems. To reduce the computational cost of feature tracking and improve the real-time performance of the robot, a lightweight SLAM method of optical flow tracking based on pyramid IMU prediction is proposed. On this basis, a centralized multi-robot SLAM system based on a robot-edge-cloud layered architecture is proposed to realize real-time collaborative SLAM. It avoids the problems of limited on-board computing resources and low execution efficiency of single robot. In this framework, only the feature points and keyframe descriptors are transmitted and lossless encoding and compression are carried out to realize real-time remote information transmission with limited bandwidth resources. This design reduces the actual bandwidth occupied in the process of data transmission, and does not cause the loss of SLAM accuracy caused by data compression. Through experimental verification on the EuRoC dataset, compared with the current most advanced local feature compression method, our method can achieve lower data volume feature transmission, and compared with the current advanced centralized multi-robot SLAM scheme, it can achieve the same or better positioning accuracy under low computational load. |
13 pages, 18 figures |
| ShapeR: Robust Conditional 3D Shape Generation from Casual Captures | 2026-01-16 | ShowRecent advances in 3D shape generation have achieved impressive results, but most existing methods rely on clean, unoccluded, and well-segmented inputs. Such conditions are rarely met in real-world scenarios. We present ShapeR, a novel approach for conditional 3D object shape generation from casually captured sequences. Given an image sequence, we leverage off-the-shelf visual-inertial SLAM, 3D detection algorithms, and vision-language models to extract, for each object, a set of sparse SLAM points, posed multi-view images, and machine-generated captions. A rectified flow transformer trained to effectively condition on these modalities then generates high-fidelity metric 3D shapes. To ensure robustness to the challenges of casually captured data, we employ a range of techniques including on-the-fly compositional augmentations, a curriculum training scheme spanning object- and scene-level datasets, and strategies to handle background clutter. Additionally, we introduce a new evaluation benchmark comprising 178 in-the-wild objects across 7 real-world scenes with geometry annotations. Experiments show that ShapeR significantly outperforms existing approaches in this challenging setting, achieving an improvement of 2.7x in Chamfer distance compared to state of the art. |
Proje...Project Page: http://facebookresearch.github.io/ShapeR Video: https://www.youtube.com/watch?v=EbY30KAA55I |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| A Tightly Coupled Magnetic-Inertial SLAM System for Field Robots | 2026-11-02 | ShowField robots operating in mines, tunnels, and disaster areas require reliable localization despite GNSS denial and degraded camera and LiDAR measurements caused by dust, smoke, or poor illumination. Missions involving both ground and aerial robots further motivate complementary sensing modalities that do not depend on visibility, preinstalled infrastructure, or wheel contact. This letter presents MISLAM, a tightly coupled magnetic-inertial simultaneous localization and mapping system using an IMU and a nine-magnetometer array, without relying on external odometry or platform-specific motion assumptions. The system combines the inertial navigation model with local polynomial magnetic-field models for short-term motion estimation and a global Gaussian process model for drift correction using an error-state Kalman filter. Offline evaluation on three indoor datasets collected with a handheld rig shows reductions in total endpoint closure error of 45.7% - 78.0% relative to magnetic-inertial odometry. Under poor illumination, MISLAM maintains tracking after visual-inertial odometry fails, achieving a horizontal closure error of 0.332 m, although vertical drift remains a limitation. These results support magnetic-inertial SLAM as a complementary localization modality for robotic sensor suites operating under degraded visual conditions. |
|
| Visual-Inertial RGB-D SLAM with Encoders for a Differential Wheeled Robot | 2026-10-30 | ShowRecent years have seen multiple impressive results in visual-inertial odometry (VIO) techniques, by which accurate state estimation can be achieved via the extended Kalman filter (EKF) or nonlinear optimization. However, these approaches are rarely open source, and they tend to fail in real experiments due to the temporary lack of feature points and fast motion. Therefore, in this study, we used encoders to overcome the temporary failure of purely vision-based simultaneous localization and mapping (SLAM). Here, we propose a generative measurement model for encoders and derive an expression for the maximum a posteriori (MAP) estimate and necessary Jacobians for optimization. We use our theory to present a novel tightly coupled visual-inertial encoder RGB-Depth (RGB-D) SLAM system. Tests on our system on an in-house dataset confirmed that our modeling effort led to accurate (with a root mean squared error (RMSE) of approximately 2-7 cm) and robust state estimation in real time. The source code and our dataset containing the encoder information have been published for verification. |
11 pa...11 pages. Author-accepted manuscript of the article published in IEEE Sensors Journal, 2022 |
| OpenVIO: A Versatile Open-Source Visual-Inertial Odometry System | 2026-10-30 | ShowVisual-inertial odometry (VIO) is a fundamental topic in robotics research. Beyond research, VIO has been adopted in a wide range of products. In this manuscript, we present OpenVIO, an open-source VIO system for research and practical applications. It is released under the MIT license. OpenVIO supports pinhole and fisheye cameras and extends this support to stitched 360° images. We describe each part of the system together with its design considerations. We also report benchmark results with other open-source VIO systems on public datasets. OpenVIO is available as open source at https://github.com/93won/openvio. |
7 pages, 6 figures |
| OpenJaws: An Open Hardware UMI Framework with Stereo Localization for Generalist Manipulation | 2026-10-27 | ShowHandheld data-collection interfaces such as the Universal Manipulation Interface (UMI) allow manipulation demonstrations to be gathered in the wild without a robot in the loop, but the quality of the resulting action labels is bounded by the accuracy of vision-based localization. Existing systems recover end-effector pose from monocular visual-inertial odometry, which is subject to scale ambiguity and drift, or delegate tracking to closed-source modules and external infrastructure that reintroduce cost and setup burden. We present OpenJaws, a fully open hardware and software framework for sensorized UMI that improves localization and simplifies deployment across robot platforms. OpenJaws pairs a modular, 3D-printed gripper exposing a standardized ISO adapter with a low-cost stereo camera to recover metric camera trajectories. We further introduce UMI-Loc-Bench-v1, a benchmark quantifying real-world localization accuracy for handheld data-collection devices, and show that stereo localization substantially outperforms monocular alternatives. Finally, we demonstrate that policies trained on data collected with OpenJaws can be deployed on a real robot across multiple tasks. Hardware, software, and data are released under a commercially permissive license. Code, data, and videos are available at https://open-jaws.github.io/ |
Proje...Project page: https://open-jaws.github.io/ |
| Event-based SLAM Benchmark for High-Speed Maneuvers | 2026-10-22 | ShowEvent-based cameras are bio-inspired sensors with pixels that independently and asynchronously respond to brightness changes at microsecond resolution, offering the potential to handle visual tasks in high-speed maneuvering scenarios. Existing event-based approaches, although successful in mitigating motion blur caused by high-speed maneuvers, suffer from many limitations. Some of them highlight a success of pose tracking for a fronto-parallel fast shaking camera closed to the structure, while others assume pure (optionally aggressive) three-degree-of-freedom rotations. The former requires persistent local map visibility within the field of view (FOV), whereas the latter fails to generalize to six-degree-of-freedom (6-DoF) motions where both linear and angular velocities may be large. Consequently, current successes do not fully demonstrate that event-based state estimation under arbitrary aggressive maneuvers is a fully solved problem. To quantitatively assess the extent to which the potential of event cameras has been unlocked, we conduct a thorough analysis of state-of-the-art (SOTA) event-based visual odometry (VO)/visual-inertial odometry (VIO) methods and report shortcomings in current public datasets. Furthermore, we introduce a benchmarking framework for event-based state estimation, called EvSLAM, characterized by sufficient variation in data collection platforms, diverse extreme lighting scenarios, and a wide scope of challenging motion patterns under a clear and rigorous definition of high-speed maneuvers for mobile robots, along with a novel evaluation metric designed to fairly assess the operational limits of event-based solutions. This framework benchmarks state-of-the-art methods, yielding insights into optimal architectures and persistent challenges. |
|
| Onboard UAV State Estimation for Fast Flights Without External Localization Systems: A2RL Drone Racing Finalist Approach | 2026-10-21 | ShowFast flights with aggressive maneuvers in cluttered GNSS-denied environments require fast, reliable, and accurate UAV state estimation. In this paper, we present an approach for onboard state estimation of a high-speed UAV using a monocular RGB camera and an IMU. Our approach fuses data from Visual-Inertial Odometry (VIO), an onboard landmark-based camera measurement system, and an IMU to produce an accurate state estimate. Using onboard measurement data, we estimate and compensate for VIO drift through a novel mathematical drift model. State-of-the-art approaches often rely on more complex hardware (e.g., stereo cameras or rangefinders) and use uncorrected drifting VIO velocities, orientation, and angular rates, leading to errors during fast maneuvers. In contrast, our method corrects all VIO states (position, orientation, linear and angular velocity), resulting in accurate state estimation even during rapid and dynamic motion. Our approach was thoroughly validated through 1600 simulations and numerous real-world experiments. Furthermore, we applied the proposed method in the A2RL Drone Racing Challenge 2025, where our team advanced to the final four out of 210 teams and earned a medal. |
Visit...Visit our webpage for more details: https://mrs.fel.cvut.cz/papers/onboard-uav-state-estimation-a2rl |
| Visual-Inertial Odometry for Ground Robots with Dynamic Camera-IMU Extrinsic Calibration | 2026-10-20 | ShowVisual-inertial odometry systems are often constrained by the assumption that sensors are rigidly mounted, which limits their adaptability in practical scenarios involving structural motion. To address this limitation, this paper investigates time-varying camera-IMU extrinsic calibration. First, a dynamic camera-IMU extrinsic calibration method based on the Denavit-Hartenberg convention is proposed by incorporating the robot kinematic model. Second, a two-step initialization strategy is developed to efficiently estimate the initial joint angles, providing a reliable starting point for subsequent extrinsic parameter optimization and enabling online estimation of timevarying camera-IMU extrinsic parameters. Simulation experiments demonstrate that the proposed framework can effectively accommodate variations in camera-IMU extrinsic parameters while maintaining competitive localization accuracy. |
|
| AnchorFlow-VIO: Fixed-Template Monocular Visual-Inertial Odometry on LaMAria | 2026-10-19 | ShowWe present AnchorFlow-VIO, an open-source Basalt-derived monocular visual-inertial odometry system, and evaluate a uniform configuration on LaMAria. The system uses one grayscale camera and one IMU, combining factory IMU correction, a fixed timing adjustment, windowed initialization and fixed-template patch tracking. Server evaluation of all 63 test recordings yields control-point scores of 47.2, 37.1, 24.8, 35.1 and 15.6 for Short, Medium, Long, Low light and Moving platform. Test pose recalls at 5 m are 84.6%, 69.1%, 53.9% and 62.9% for the four categories reporting dense recall. On ten training recordings, mean control-point score is 41.38 and all 265,694 image timestamps receive poses, including 204 filled poses. One training platform recording diverges and scores zero. A paired four-record comparison favors fixed over rolling templates on three recordings, while reversing that ranking on the fourth. Increasing feature density also yields mixed results. Ten fresh runs at prospectively fixed 30 s and 120 s offsets expose substantial start sensitivity: a low-light recording scores 6.14 at the earlier start and 42.00 at the later start. These results characterize the limits of the frozen system and show why complete output, feature counts and selected successful runs are insufficient evidence of robust localization. Source code, frozen configuration and reproducible submission tools are publicly available; training ablations are reported separately from the submitted test evaluation. |
14 pa...14 pages, 5 figures, 8 tables. Code: https://github.com/sudhirpratapyadav/anchorflow-vio |
| Mono-Hydra++: Real-Time Monocular Scene Graph Construction with Multi-Task Learning for 3D Indoor Mapping | 2026-10-19 | ShowAutonomous agile robots need more than metric geometry: they must understand objects, rooms, places, and spatial relations for search, inspection, exploration, and human-robot interaction. Conventional metric maps support localization and collision avoidance, but do not provide this semantic and relational structure. 3D scene graphs address this gap by connecting geometry with object-level and room-level understanding. Building such representations on agile platforms remains difficult because aerial and lightweight robots operate under strict payload, power, and compute limits, making RGB-D cameras and LiDAR sensors impractical for many onboard settings. We present Mono-Hydra++, a real-time monocular RGB plus IMU pipeline for indoor metric-semantic mapping and hierarchical 3D scene graph construction. The system combines M2H-MX, a DINOv3-based multi-task model for depth and semantics, with a deep-feature visual-inertial odometry front end, sparse predicted depth constraints in the VIO-derived pose graph, semantic masking for dynamic regions, and pose-aware temporal alignment before volumetric fusion in the Mono-Hydra backend. On the Go-SLAM ScanNet evaluation subset, Mono-Hydra++ achieves 1.6% lower average trajectory error than the strongest RGB-D baseline in our comparison, while using only monocular RGB plus IMU input. On calibrated 7-Scenes, it improves average ATE by 29.8% over the strongest competing calibrated baseline. We further validate Mono-Hydra++ in a real ITC building deployment using RealSense RGB plus IMU and demonstrate embedded feasibility by deploying the ONNX/TensorRT FP16 M2H-MX-L perception model at 25.53 FPS on a Jetson Orin NX 16GB. These results show that Mono-Hydra++ can provide real-time metric-semantic mapping and scene graph construction for resource-constrained robotic platforms without relying on active depth sensors. |
Submi...Submitted to ISPRS Journal of Photogrammetry and Remote Sensing. 50 pages, figures and tables included. Code: https://github.com/BavanthaU/mono-hydra-pp.git |
| FoundationVIO: Enforcing Metric Scale in Feed-forward Visual-Inertial Odometry | 2026-10-10 | ShowFeed-forward geometric foundation models (GFMs) directly predict dense geometry and camera poses from batches of images. Emerging visual SLAM methods chain together consecutive GFM batches to form dense point cloud maps. A core challenge is how to address the ambiguous geometric scale of GFM predictions: independent predictions are locally consistent up-to-scale but not constrained by any direct physical measurements. Models trained on data with known scale improve the consistency of GFM reconstructions but again do not directly measure it. Consequently, trajectories and 3D reconstructions from GFM-enabled SLAM methods inherit this issue. Motivated by parallels with classical visual-inertial odometry, we introduce FoundationVIO which integrates inertial sensing with GFM predictions to recover metric scale reconstruction using a factor-graph based approach. We demonstrate odometry estimates competitive with classical state-of-the-art VIO methods and accurate dense reconstructions that outperform GFM-enabled SLAM methods. We also add a minimal GFM-based loop-closure mechanism to create a full Visual-Inertial SLAM system capable of creating globally accurate reconstructions and pose estimates. We evaluate our method on three datasets with well known sensor systems and demonstrate its compatibility on ubiquitous camera and IMU sensors. The method runs at approximately real-time on a consumer-grade laptop GPU. Code: https://github.com/ori-drs/FoundationVIO. |
|
| Reference-Free Food Volume Estimation from Egocentric Stereo Video | 2026-10-08 | ShowCurrent stereo vision-based approaches for food portion estimation are hindered by the user burden of using a mobile phone to capture multiple images or placing a reference object in the scene for scale calibration. As a step towards passive dietary intake monitoring, we introduce a method that uses stereo video from a wearable egocentric device to reconstruct food items in true physical scale to accurately estimate volume. By utilizing FoundationStereo for zero-shot stereo matching, SAM 2 for food segmentation, and TSDF fusion over a set of keyframes, our method yields a visually and metrically accurate 3D mesh of any plated food item from a single stereo recording. We leverage the unique offerings of wearables, such as a factory-calibrated stereo baseline and camera poses estimated on-device via visual-inertial odometry (VIO), to perform fully automatic reconstruction without a scale reference object in the scene. We show strong initial volume estimation results across five recorded food items, achieving a mean absolute percentage error of 9.1%. |
9 pag...9 pages, 4 figures, 1 table |
| Probabilistic Pedestrian Forecasts from a Handheld Phone: World-Frame Heat Maps, Visual-Inertial Height Drift, and Evaluation without Ground Truth | 2026-10-03 | ShowA pedestrian with a phone could be shown where nearby people will be in the next few seconds, if the forecast stays on the ground while the phone moves, is calibrated, and needs only a monocular camera and visual-inertial odometry (VIO). We build and evaluate such a system. People are detected, lifted onto the floor by ray-plane intersection, tracked in a gravity-aligned metric frame, and forecast as per-step probability maps by a small U-Net trained with a negative log-likelihood (NLL) loss on bird's-eye (SDD) and first-person (EgoTraj-Bench) trajectories. On handheld ADVIO recordings, vertical VIO drift and the user's own changes of level silently rescale monocular ground positions (by 87% within 90 s on one clip; on another, all tracks are lost for the last 31% of the clip); keeping the camera's height above the floor constant under a low-pass-filtered altitude avoids this, though it lags on escalators. On the SDD and EgoTraj-Bench test splits, the final forecaster lowers the NLL at 4.8 s by 1.51 and 1.37 nats relative to a fitted constant-velocity Gaussian. Lacking ground truth for people in handheld video, we score forecasts against the tracker's own later raw measurements. In an internally pre-registered evaluation on seven held-out clips, the final forecaster's NLL is lower than the benchmark-fitted baseline's at 1.2, 2.4 and 4.8 s (by 0.17, 0.23 and 0.47 nats; 95% intervals over people exclude zero), but by less than half as much as on the development clips. Exploratory analyses cut both ways: resampling clips instead of people widens the intervals to include zero at 1.2 and 2.4 s, and once both forecasters are recalibrated on the development clips the network is significantly better only at 1.2 s; but two clips run with ADVIO's reference poses favour the network much more when re-run with the phone's own poses. We discuss what such self-consistency scores can and cannot show. |
28 pa...28 pages, 7 figures, 16 tables |
| HexVIO: Towards All-Day Stereo-Inertial Tracking Through Commodity DSPs | 2026-10-02 | ShowThe ability of a device to localize itself within its surroundings is a fundamental prerequisite for spatial computing. Visual-inertial odometry (VIO) has proven to be a cost-effective and accurate solution for this task. Robots, wearables, XR devices, and drones can benefit significantly from efficient implementations of VIO since they allow for cooler, lighter, and cheaper devices with longer battery life and a better user experience. In this work, we propose to enhance the efficiency of a VIO system by leveraging the Hexagon DSP, a commodity co-processor present in many modern smartphones and XR devices. Our approach offloads the visual frontend of a stereo-inertial odometry system to the DSP while keeping the backend on the main CPU. By optimizing the implementation for the DSP architecture, we achieve significant reductions in power consumption and latency compared to CPU-only execution. Our system, HexVIO, demonstrates a 67% reduction in power consumption or an 86% increase in throughput on a commodity smartphone, with the ability to sustain long-term real-time 30 fps tracking for 0.83 W, corresponding to ~18 hours of tracking on the testing device. These results highlight the potential of commodity DSPs for enabling all-day visual-inertial tracking in robotics and mobile devices. |
|
| LBDU-VIO: Learned Bias Dynamics and Uncertainty for Visual-Inertial Odometry with Unreliable Vision | 2026-09-30 | ShowVisual-inertial odometry (VIO) for aerial robots relies on high rate inertial measurement unit (IMU) propagation between visual updates. However, conventional multi state constraint Kalman filters (MSCKFs) use random walk bias assumptions and fixed noise parameters, which can limit robustness when visual information is unreliable. To address this problem, we propose LBDU-VIO, a learning-augmented MSCKF with learned continuous time bias dynamics and an IMU uncertainty model. A neural ordinary differential equation (ODE) models continuous time bias dynamics to propagate the filter's bias states, replacing their random walk model. The IMU uncertainty model predicts motion adaptive measurement noise covariances for covariance propagation. Both models are trained with pose supervision without direct labels. Experiments on real world EuRoC and TUM-VI benchmarks show lower errors than representative visual-inertial baselines, including a 25.1% reduction in mean relative position error compared with S-MSCKF on EuRoC sequences with 10s visual outage. |
12 pages, 8 figures |
| VkVIO: Cross-platform GPU Acceleration for Visual-Inertial Odometry with Vulkan | 2026-09-24 | ShowPerception in robotics and XR fundamentally relies on good state estimation. Visual-inertial odometry (VIO) and Simultaneous Localization and Mapping (VI-SLAM) are proven ways of achieving this goal in a cost-effective and accurate manner. Efficiency in these systems allows for smaller, cooler, and lighter devices. GPU acceleration is a natural approach for reducing latency, thanks to their wide availability in platforms like embedded computers, mobile phones, and XR headsets. However, previous works in the literature have limited themselves to the use of CUDA for this task, significantly reducing deployment options to a single vendor. We instead leverage the vendor-agnostic Vulkan API, originally designed for the strict performance requirements of 3D graphics applications. In this work, we present VkVIO, the first, to the best of our knowledge, cross-platform GPU-accelerated VIO method. We provide state-of-the-art accuracy with causal estimates required for real-time operation. We deploy VkVIO on a diverse range of devices spanning a workstation, a laptop, and an extremely inexpensive single-board computer, while outperforming CUDA-based systems on the same hardware. VkVIO enables possibilities for low-latency, low-power, and low-cost VIO in robotics and XR. |
|
| TinyCVIO: A Constellation-Aided Visual-Inertial Odometry System for Nanodrones | 2026-09-24 | ShowNanodrones require accurate, real-time state estimation under severe sensing and computational constraints. We present TinyCVIO, a visual-inertial odometry system that co-designs miniature sensing, visual processing, and estimation for a commodity dual-core microcontroller with 520 kB SRAM. Lightweight LED constellations provide known geometry without surveyed positions or yaw angles, assuming placement on a common level plane. A streaming visual frontend tracks LED observations from a millimeter-scale camera at 29.2 FPS, while a rigid-board measurement model retains inter-LED constraints and streaming QR bounds estimation workspace for a fixed filter-state size. Across 19 hand-held hardware-in-the-loop datasets, the rigid-board model reduces mean absolute trajectory error by 27% relative to planar points. The complete system runs onboard a Crazyflie across nine flights at three speeds, achieving 3.5-3.7 cm mean absolute trajectory error and 0.50-0.60% relative pose error over 10 m segments, with mean estimate latency of 15.7-16.3 ms. |
9 pag...9 pages, 6 figures, 4 tables. Video: https://youtu.be/6hQNdIcjHJE |
| MDE-VIO: Enhancing Visual-Inertial Odometry Using Learned Depth Priors | 2026-09-24 | ShowTraditional monocular Visual-Inertial Odometry (VIO) systems struggle in low-texture environments where sparse visual features are insufficient for accurate pose estimation. To address this, dense Monocular Depth Estimation (MDE) has been widely explored as a complementary information source. While recent Vision Transformer (ViT) based complex foundational models offer dense, geometrically consistent depth, their computational demands typically preclude them from real-time edge deployment. Our work bridges this gap by integrating learned depth priors directly into the VINS-Mono optimization backend. We propose a novel framework that enforces affine-invariant depth consistency and pairwise ordinal constraints, explicitly filtering unstable artifacts via variance-based gating. This approach strictly adheres to the computational limits of edge devices while robustly recovering metric scale. Extensive experiments on the TartanGround and M3ED datasets demonstrate that our method prevents divergence in challenging scenarios and delivers significant accuracy gains, reducing Absolute Trajectory Error (ATE) by up to 28.3%. Code will be made available. |
6 pag...6 pages, 2 figures, 3 tables. Submitted to ICIP 2026 |
| Robust Structureless Monocular Visual Inertial Initialization Exploiting Line Features and Vanishing Points | 2026-09-18 | ShowAccurate initialization is essential for reliable visual-inertial odometry (VIO), but it is often ill-conditioned under degenerate motions. Existing methods typically require restrictive excitation motions to ensure sufficient observability or rely on computationally expensive 3D structure reconstruction, limiting efficient and practical deployment. To address these limitations, we propose SLIM-init, a structureless monocular VIO initializer that directly exploits geometric constraints from tracked 2D line features without explicit 3D landmark reconstruction. Specifically, SLIM-init leverages line-derived vanishing points (VPs) as translation-invariant orientation cues to provide robust rotation-only constraints under degenerate scenarios such as low-parallax or translation-dominant motions. It further incorporates a line epipolar residual to constrain translation and a line-normal projection residual to improve the conditioning of linear alignment, enhancing the accuracy and robustness of initial state estimation. Extensive experiments on a public benchmark and challenging custom degenerate-motion sequences demonstrate improved accuracy and robustness over state-of-the-art initialization methods. The source code is available at: https://github.com/cjunwan/SLIM-init. |
8 pag...8 pages, 5 figures, Accepted to IROS 2026 |
| NSFlow: End-to-End Differentiable Neuro-Symbolic Optical Flow for Visual Odometry | 2026-09-16 | ShowSparse optical flow provides stable inter-frame correspondence, playing a key role in Visual Odometry (VO) and Visual-Inertial Odometry (VIO). Classical optimization-based methods, such as Lucas-Kanade (LK), perform well under small displacements but are sensitive to large motions and illumination changes. Modern regression-based learning methods, while more robust in complex scenes, are often computationally heavy and lack explicit geometric consistency, making them less suitable for efficient VO/VIO front-ends. To bridge this gap, we propose a hybrid neuro-symbolic framework that combines the strengths of both paradigms. Our method uses a Convolutional Neural Network (CNN) to extract robust feature representations, which is fed into a differentiable LK optimizer to estimate optical flow in an end-to-end trainable manner. Through implicit differentiation, gradients are propagated across the iterative solver, enabling joint optimization of feature extraction and flow estimation. The resulting system integrates seamlessly into existing VO/VIO pipelines and runs in real-time on embedded platforms. Experiments show that our method outperforms conventional optimization-based flow in challenging conditions such as dynamic lighting and low texture, while also achieving higher accuracy and lower latency than purely regression-based alternatives. When deployed in a VIO system, our method demonstrates significant performance improvement, achieving an average error reduction of 42% on challenging datasets while enhancing tracking stability. The code is publicly available. |
14 pages, 9 figures |
| Benchmarking Visual-Inertial Odometry in Subterranean Environments Under Sensor Degradation, Miscalibration, and Dynamic Occlusion | 2026-09-16 | ShowVisual-inertial odometry (VIO) is a core capability for autonomous operation in GPS-denied subterranean environments, yet its reliability can degrade sharply under sensor drift, calibration errors, and dynamic occlusion. Existing evaluations mainly emphasize nominal-condition accuracy, offering limited insight into when practical deployment failures occur. In this work, we present a failure-centric stress-test benchmark for VIO in underground environments using the CERBERUS dataset. We systematically evaluate four representative VIO systems spanning filtering-, optimization-, and learning-based paradigms under nine practical perturbation settings, including IMU bias and noise variation, camera intrinsic and extrinsic drift, and dynamic scene occlusion. Beyond conventional trajectory error, we analyze robustness limits through coverage ratio and failure thresholds, revealing breakdown behaviors that are not captured by nominal-condition performance alone. Our study shows distinct vulnerability patterns across VIO paradigms: some methods are more sensitive to inertial degradation, while others are more affected by geometric miscalibration or dynamic interference. These results provide deployment-oriented guidance for VIO selection, calibration prioritization, and reliable operation in challenging underground scenarios. To support reproducible evaluation and future extensions, we will release the full benchmark scripts and evaluation pipeline. |
8 pages, 5 figures |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| Smartphone Indoor Localization Enhanced by Hierarchical Multi-Sensor Context Recognition | 2026-10-29 | ShowSmartphone indoor localization requires low-cost, robust positioning across scene transitions and phone-use variations. Pedestrian dead reckoning (PDR) enables infrastructure-free deployment but accumulates errors from heading drift, structural transitions, and carrying-mode changes. Existing methods often treat contextual cues as a single event stream or local correction signal, limiting their ability to address distinct localization errors. This paper presents a framework enhanced by hierarchical multi-sensor context recognition. A jointly data- and model-driven recognition model predicts environment, scene, and activity probabilities, which are converted into scene-event and action-context segments. Scene-event segments guide scene-constrained PDR correction through corridor heading constraints, landmark attraction, fixed-lag backtracking, and vertical-transition handling. Action-context segments drive action-aware RoNIN-based position compensation for carrying-mode-dependent drift. Experiments on cross-floor and multi-action trajectories against LiDAR-SLAM references show that the proposed framework outperforms PDR and representative learned inertial odometry baselines, demonstrating that hierarchical semantics provide reliable, localization-specific correction cues. |
|
| SKiD-SLAM: Resource-Aware Distributed Multi-Robot LiDAR SLAM for Planetary Missions with Field Deployments at Analog Sites | 2026-10-28 | ShowFuture planetary exploration missions require teams of robots to collaboratively map unknown environments under severe communication limits, constrained onboard computation, and no central infrastructure. In this field report, we present SKiD-SLAM, a distributed multi-robot LiDAR simultaneous localization and mapping (SLAM) system for such environments, and report its deployment across two mission concepts: cave exploration and scientific surface exploration. We survey the communication and computation budgets of actual planetary missions and select modules that fit these budgets for inter-robot place recognition, relative pose estimation, and distributed trajectory fusion. Before field deployment, we verify the assembled system on public datasets resembling planetary environments and in a simulator built on NASA JPL's Open Source Rover. We then validate SKiD-SLAM through field deployments in South Korea: a team of wheeled rovers explores the Majang lava-tube cave on Jeju Island, and quadruped robots carry out a crater-and-rock science scenario on a planetary analog terrain at the Korea Institute of Civil Engineering and Building Technology (KICT) in Yeoncheon. Across both deployments, SKiD-SLAM reliably performed distributed multi-robot LiDAR SLAM within a 250 kbps inter-robot communication budget, demonstrating its feasibility under challenging field conditions. Supplementary materials, including code and field data, are available at https://sparolab.github.io/research/skid_slam/. |
31 pa...31 pages. Published in IEEE Transactions on Field Robotics. Code and field data: https://sparolab.github.io/research/skid_slam/ |
| PinNet: Keypoint-Aware Learned Local Descriptors with Geometric Embedding for Loop Closure in LiDAR SLAM | 2026-10-16 | ShowLoop closure is essential to reduce drift and build globally consistent maps in large-scale environments. However, reliable loop closure with only geometric information from, e.g., a LiDAR sensor, remains challenging due to the difficulty of constructing discriminative geometric features. We present PinNet, a neural network that produces local geometric descriptors from point clouds for place recognition and scanto-scan registration. PinNet incorporates a neural network that generates keypoints and their corresponding descriptors, together with a plane-based geometric self-attention module that models inter-keypoint spatial relationships to enhance descriptor discriminability for loop-closure detection and point-cloud registration. The approach is comprehensively evaluated on multiple datasets collected with different LiDAR sensors. Experimental results demonstrate strong place-recognition performance, precise relative pose estimation, and successful single-shot localization in different environments. |
fixed...fixed LCDNet registration errors |
| RTISSLAM-3D: Full 6-DoF SLAM With an In-Air 3D Imaging Sonar | 2026-10-15 | ShowEcholocating bats navigate dense forests using sonar alone, and in-air 3D imaging sonars now capture the entire frontal hemisphere with every pulse. Sonar SLAM, however, has so far been planar: current systems estimate a 2D trajectory and discard the elevation information in the sonar images, an assumption that does not hold on rough and sloped terrain such as a forest floor. In this paper, we introduce RTISSLAM-3D, a novel 6-DoF SLAM system for a wheeled robot that carries an in-air 3D imaging sonar, an IMU and wheel encoders. A CFAR detector on the sphere reduces every sonar image to a sparse set of 3D detections, which we use for a 3D acoustic odometry that models the shape of the sonar beam, for place recognition with polar submap descriptors, and for the verification of loop closures by 3D registration, all combined in a 6-DoF pose graph with attitude factors. We evaluated RTISSLAM-3D on a 545m drive through a forest, with a 3D lidar SLAM as reference. The loop closures reduced the horizontal error of the acoustic odometry from 1.74m to 0.28m, with a height error of 0.12m, and 94.6% of the resulting 3D sonar map lies within 0.3 m of the lidar map. |
|
| TAPVid-MV: A Benchmark for Tracking Any Point in 3D Across Multiple Views | 2026-10-09 | ShowMulti-camera systems are increasingly practical for robotics, AR/VR, and autonomous driving because complementary views reduce depth ambiguity and preserve visibility under occlusion. Existing point-tracking benchmarks, however, focus on a single video or static multi-camera rigs. None test long-term 3D point tracking across several synchronized views under camera motion. We introduce TAPVid-MV (Tracking Any Point in Video across Multiple Views), the first benchmark for this setting. It contains a curated set of 238 sequences, 1,004 calibrated camera streams, and 74,147 point tracks across seven subsets spanning indoor and outdoor domains, from robotics and human activity to driving and synthetic procedural scenes. We obtain these trajectories using dataset-specific auxiliary modalities: sensor depth, LiDAR, SLAM and SfM points, human meshes, posed object meshes, and simulation. Every sequence and trajectory is visually verified by human annotators. Across more than 30 baselines, no method comes close to solving the task. Surprisingly, existing multi-view point trackers do not consistently outperform monocular point trackers. By evaluating reconstruction and point tracking on the same datasets, TAPVid-MV helps distinguish errors in recovered geometry from errors in point correspondence. Through this joint analysis, we identify geometry recovery as a major bottleneck for accurate 3D point tracking. Beyond multi-view 3D point tracking, our released annotations support monocular 2D and 3D point tracking, future-trajectory prediction, and 4D reconstruction. Project page: https://tapvidmv.github.io/ |
|
| Degeneracy-Orthogonal Geometric Constraints for LiDAR SLAM | 2026-09-29 | ShowAutonomous robot navigation relies on simultaneous localization and mapping (SLAM) to estimate motion and maintain an accurate pose within an environment. However, in axially uniform corridors such as long tunnels and pipelines, LiDAR odometry is fundamentally limited by unconstrained drift along the feature-weak travel direction. This structural degeneracy cannot be resolved by local scan matching alone. To address this challenge, we propose the Degeneracy-orthogonal Contour Offset Descriptor (DeCOD), a structure-aligned geometric descriptor for cross-sectional landmarks. Cross-sectional boundaries, such as pipe joints and structural rings, provide metric constraints along this degenerate axis, but distinguishing individual landmarks requires capturing subtle surface variations across nearly identical profiles. The descriptor parameterizes signed normal deviation from estimated boundary contours, and matching explicitly resolves heading ambiguity and decouples first-order contour errors by distortion estimation. Matched landmarks yield geometric factors that enforce agreement in cross-section position and corridor axis alignment during pose-graph optimization, correcting longitudinal drift while leaving rotation about the common axis unconstrained. On a public benchmark and in field experiments, DeCOD achieves robust landmark retrieval over standard 3D descriptors and successfully stabilizes trajectories across different odometry frontends, reliably constraining longitudinal drift under geometric degeneracy. |
8 pages, 4 figures |
| Parameter Sensitivity Analysis for Aerial LiDAR-Inertial Odometries in low-altitude flights | 2026-09-11 | ShowLiDAR-based SLAM (Simultaneous Localization and Mapping) and LIO (LiDAR-inertial odometry) algorithms are often used for precise navigation of unmanned aerial vehicles, especially during interactions with the aerial robot's environment. However, the performance of these algorithms is greatly dependent on the scenario, LiDAR, and robot motion characteristics, often requiring an intensive tuning process to achieve the desired performance. To aid these tuning efforts, this paper analyzes the influence on performance of the parameters of an EKF-based LIO algorithm (FAST-LIO2) and the LIO module of a graph-based SLAM algorithm (Cartographer) on aerial LiDAR SLAM datasets recorded using different LiDARs in low-to-moderate-altitude flights in diverse environments. The analysis is conducted on the absolute trajectory error (ATE) resulting from processing the datasets with the LIO algorithms configured with each combination of parameters obtained in an exhaustive grid search. The relationship between individual parameters and the ATE results is assessed using Pearson's correlation, while the influence of each parameter is assessed using random forest permutation importance analyses with random forest models trained to predict the resulting ATE values based on the choice of parameters. The performed analysis obtains for Cartographer and FAST-LIO2: i) the identification of parameters with stronger influence in performance, ii) a simplified tuning procedure, and iii) tuning recommendations. Using the proposed tuning recommendations, both algorithms obtain on the analyzed datasets ATE values within 5 cm to the optimal performance found in the grid search procedure in 94% of the analyzed cases. |
19 pa...19 pages, 5 figures, 14 tables |
| Chain-SLAM: Globally Consistent Backend for Multi-Session LiDAR SLAM via Chained Loop Closure | 2026-09-10 | ShowMaintaining consistency over long spatial and temporal horizons remains a fundamental challenge in large-scale LiDAR SLAM, particularly when integrating maps collected across multiple sessions. We present Chain-SLAM, a LiDAR SLAM backend enabling online multi-session map alignment and reuse with global consistency at large scale. We implement a chained loop closure mechanism that efficiently propagates geometric constraints across inter-session keyframes through an adjacency graph, enabling robust long-horizon consistency triggered by reliable short-horizon loop closures. The system initializes inter-session alignment with GNSS-proximity place recognition, then performs on-the-fly loop closure detections and joint optimization of loaded maps and newly acquired trajectories within a unified factor graph, maintaining both inter- and intra-session geometric consistency without dynamic object removal, and cross-platform robustness with minimal hyperparameter tuning. Experimental results show improved trajectory accuracy and robust multi-session integration on large-scale datasets. We release our source code to support reproducible research in large-scale multi-session LiDAR SLAM. Project site: https://ai4ce.github.io/Chain-SLAM/ |
Accep...Accepted to IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026. 8 pages, 6 figures |
| A comparative study on the accuracy & repeatability of mobile robotic platforms for the delivery of precision NDE measurement | 2026-09-03 | ShowMobile robotic platforms offer a flexible alternative to fixed manipulators for non-destructive evaluation (NDE) of large aerospace structures, but their base-positioning accuracy and how that accuracy should inform deployment have not been assessed under a common, externally referenced protocol. This work presents a laser tracker-based evaluation workflow (ground truth approximately 6 micrometers) that measures the static and segmented trajectory positioning accuracy of five commercial mobile platforms (KUKA KMP-1500, KUKA KMR, MiR250, Boston Dynamics Spot, Clearpath Husky) under a common protocol. A coupled multi-corner calibration recovers the laser-to-robot transformation and reflector offsets; ordinary least squares over all poses is used, with robust estimation retained only as a blunder check. Static positioning accuracy ranged from a median of 8.2 mm (KMP-1500) to 63.5 mm (Spot), with the wheel-odometry-only Husky uncalibratable. Dynamic path following was characterised by cross-track error; the component was insensitive to temporal alignment, which ranged from 6.9 mm (KMP-1500) to 112.1 mm (Spot). Both accuracy and calibratability tracked localisation capability, from the newest LiDAR SLAM platform to map-free visual odometry. No configuration meets the 0.2 to 1.0 mm aerospace NDE tolerance from the base alone; the results are framed as a design input that sizes the supplementary sensing each platform requires: roughly one order of magnitude for the best platform and nearly two for the worst, providing a reproducible basis for platform selection rather than a feasibility claim. |
|
| HP2-SLAM: Adaptive Hybrid ICP for Robust and Efficient LiDAR SLAM | 2026-08-15 | ShowAchieving robustness, accuracy, and efficiency simultaneously remains a central challenge in light detection and ranging (LiDAR) simultaneous localization and mapping (SLAM). While learning-based approaches deliver strong benchmark performance, they often require extensive training, substantial computational resources, and struggle to generalize to unseen or degenerate environments. Geometry-based methods are efficient and interpretable, yet their performance degrades in planar or repetitive scenes due to limitations of standard iterative closest point (ICP) formulations. We present HP2-SLAM, a minimalist yet robust LiDAR SLAM framework built around a neighborhood-size adaptive hybrid ICP. Our key insight is a planarity-aware adaptive threshold that dynamically classifies correspondences based on local geometric structure and density, thereby enabling a principled balance between point-to-plane and point-to-point residuals. This formulation stabilizes alignment in both structured and degenerate environments without feature engineering, learning modules, or dataset-specific tuning. Integrated into a complete SLAM pipeline with submap management, loop closure detection, and pose graph optimization, HP2-SLAM consistently outperforms strong geometry-based baselines across publicly available datasets while maintaining real-time performance on commodity hardware. Our results demonstrate that carefully designed geometric adaptation can achieve strong generalization and robustness without sacrificing simplicity or efficiency. |
|
| Improving Map Consistency in Graph-Based LiDAR SLAM Through Information-Aware Odometry and Retroactive Loop Closure | 2026-07-15 | ShowHigh-quality maps are fundamental for robotics tasks such as navigation and planning. Although modern graph-based LiDAR SLAM systems achieve good trajectory accuracies, a low trajectory error alone does not guarantee geometrically consistent maps, particularly at revisit locations where missed loop closures and residual drift can produce local misalignments. In this work, we address the problem of jointly improving global trajectory estimation and local map quality in 3D LiDAR SLAM. We first propose a framework to efficiently estimate geometry-dependent information matrices for ICP, enabling principled weighting of odometry constraints in a pose graph. We then introduce a hierarchical loop-closure module that decouples place recognition from geometric registration, together with a retroactive loop-closure module that exploits the optimized pose graph to recover missed loop closures. We also propose an evaluation protocol to measure map consistency at revisit locations. We evaluate our SLAM system on several datasets against state-of-the-art LiDAR SLAM systems. Experimental results demonstrate global trajectory accuracies on par with or better than existing methods while consistently improving local geometric map consistency at revisit locations. These results suggest that coupling uncertainty-aware odometry with geometry-guided loop-closure refinement leads to more accurate trajectories and higher-quality maps. |
8 pag...8 pages, 4 figures, 2 tables |
| Self-supervised Geometry Reasoning for LiDAR Simultaneous Localization and Mapping | 2026-06-29 | ShowLiDAR simultaneous localization and mapping (SLAM) relies on local geometric quantities such as covariances, correspondences, and surface structures. However, most existing pipelines rely on hand-crafted estimates of local geometry and use them as fixed inputs to LiDAR SLAM, which can make the estimated local geometry noisy and unstable in sparse regions of a point cloud or when using low-resolution LiDAR. To address this issue, this paper introduces a self-supervised framework that learns an explicit symbolic representation of local geometry and uses it to improve LiDAR SLAM recursively. Specifically, each point is represented as a Gaussian distribution, allowing local geometry to be described by a covariance. Without dense geometry labels or ground-truth poses, the framework learns by maximizing the likelihood of local geometry, with self-supervision derived from consistency relations over symbolic geometric representations, including predicted covariances, correspondences, and trajectory from SLAM. The learned geometry is then fed back into LiDAR SLAM, forming a reciprocal loop in which improved geometry enhances localization and mapping, and improved localization provides cleaner supervision for subsequent geometry reasoning. This framework is backend-agnostic and can be plugged into existing LiDAR SLAM pipelines without architectural changes. Experiments on KITTI under varying LiDAR resolutions show that the proposed method improves both odometry and global registration. |
|
| Fast and Accurate Outlier-Aware LiDAR Super-Resolution for SLAM Applications | 2026-06-26 | ShowThis work tackles the challenge of enhancing low-resolution LiDAR sensors for SLAM applications through a novel Deep Unrolling-based Super-Resolution (SR) model. We integrate an outlier removal module to ensure structural integrity while maintaining real-time performance. By leveraging a model-based optimization approach, our method efficiently reconstructs high-resolution point clouds while minimizing computational overhead. The proposed SR model is evaluated within a LiDAR SLAM framework, demonstrating significant improvements in pose estimation accuracy and efficiency compared to state-of-the-art SR methods. |
6 pages, 2 figures |
| Autonomous Navigation System for Library Service Robot Based on Unitree Go2 Edu | 2026-06-02 | ShowLibraries require autonomous robots to move quietly through narrow aisles while remaining safe around readers, chairs, bags, and carts. This paper presents a ROS 2 navigation system for a Unitree Go2 Edu quadruped equipped with a 4D LiDAR, a front depth camera, and an IMU. Rather than assuming the library is rough terrain, we target the practical mobility discontinuities of real deployments, including floor transitions, temporary clutter, and partially blocked passages where low-clearance wheeled platforms are less tolerant. RTAB-Map is used for visual-LiDAR SLAM, AMCL and EKF-based sensor fusion provide localization, and a Nav2 stack with A* and DWA supports planning and local avoidance. In a real library, the system achieves 100%, 96%, and 88% success rates in static, low-density dynamic, and high-density dynamic scenes, while map validation against surveyed control distances yields a mean metric error of 3.7 cm. |
6 pag...6 pages, 5 figures, 4 tables. Accepted by WCCIS 2026 |
| A Decentralized LiDAR-SLAM System with Certifiably Optimal Pose Graph Optimization | 2026-06-02 | ShowDecentralized multi-robot LiDAR-SLAM is essential for collaborative missions but faces significant challenges in maintaining global consistency. Existing frameworks predominantly rely on local-search optimization or one-time coordinate alignment, which are prone to suboptimal convergence and long-term inconsistency, especially in large-scale or degenerate environments. To address these limitations, this paper presents the first decentralized LiDAR-SLAM system that integrates a state-of-the-art certifiably optimal Pose Graph Optimization (PGO) backend. By leveraging the Riemannian Block Coordinate Descent (RBCD) algorithm, our system ensures globally consistent trajectory estimation without requiring accurate initial guesses. Experimental results demonstrate that the proposed framework achieves superior robustness, improving trajectory RMSE by up to 48.9% compared to the state-of-the-art DiSCo-SLAM. |
In Pr...In Proceedings of the IEEE International Conference on Robotics & Automation (ICRA'26) 1st Workshop on Robot Meets GNSS and Ranging for Seamless Autonomy, Vienna, Austria, Jun. 5, 2026 |
| 3D Modeling and Automated Measurement of Concrete Cracks via Segment Anything Refinement and Visual Inertial LiDAR Fusion | 2026-05-18 | ShowVisual-Spatial Systems has become increasingly essential in concrete crack inspection. However, existing methods often lacks adaptability to diverse scenarios, exhibits limited robustness in image-based approaches, and struggles with curved or complex geometries. To address these limitations, an innovative framework for two-dimensional (2D) crack detection, three-dimensional (3D) reconstruction, and 3D automatic crack measurement was proposed by integrating computer vision technologies and multi-modal Simultaneous localization and mapping (SLAM) in this study. Firstly, building on a base DeepLabv3+ segmentation model, and incorporating specific refinements utilizing foundation model Segment Anything Model (SAM), we developed a crack segmentation method with strong generalization across unfamiliar scenarios, enabling the generation of precise 2D crack masks. To enhance the accuracy and robustness of 3D reconstruction, Light Detection and Ranging (LiDAR) point clouds were utilized together with image data and segmentation masks. By leveraging both image- and LiDAR-SLAM, we developed a multi-frame and multi-modal fusion framework that produces dense, colorized point clouds, effectively capturing crack semantics at a 3D real-world scale. Furthermore, the crack geometric attributions were measured automatically and directly within 3D dense point cloud space, surpassing the limitations of conventional 2D image-based measurements. This advancement makes the method suitable for structural components with curved and complex 3D geometries. Experimental results across various concrete structures highlight the significant improvements and unique advantages of the proposed method, demonstrating its effectiveness, accuracy, and robustness in real-world applications. |
Title...Title and author list updated |
| D-SLAMSpoof: An Environment-Agnostic LiDAR Spoofing Attack using Dynamic Point Cloud Injection | 2026-03-11 | ShowIn this work, we introduce Dynamic SLAMSpoof (D-SLAMSpoof), a novel attack that compromises LiDAR SLAM even in feature-rich environments. The attack leverages LiDAR spoofing, which injects spurious measurements into LiDAR scans through external laser interference. By designing both spatial injection shapes and temporally coordinated dynamic injection patterns guided by scan-matching principles, D-SLAMSpoof significantly improves attack success rates in real-world, feature-rich environments such as urban areas and indoor spaces, where conventional LiDAR spoofing methods often fail. Furthermore, we propose a practical defense method, ISD-SLAM, that relies solely on inertial dead reckoning signals commonly available in autonomous systems. We demonstrate that ISD-SLAM accurately detects LiDAR spoofing attacks, including D-SLAMSpoof, and effectively mitigates the resulting position drift. Our findings expose inherent vulnerabilities in LiDAR-based SLAM and introduce the first practical defense against LiDAR-based SLAM spoofing using only standard onboard sensors, providing critical insights for improving the security and reliability of autonomous systems. |
|
| MirrorDrift: Actuated Mirror-Based Attacks on LiDAR SLAM | 2026-03-11 | ShowLiDAR SLAM provides high-accuracy localization but is fragile to point-cloud corruption because scan matching assumes geometric consistency. Prior physical attacks on LiDAR SLAM largely rely on LiDAR spoofing via external signal injection, which requires sensor-specific timing knowledge and is increasingly mitigated by modern defense mechanisms such as timing obfuscation and injection rejection. In this work, we show that specular reflection offers an injection-free alternative and demonstrate an attack, MirrorDrift, that uses an actuated planar mirror to cause ghost points in LiDAR scans and systematically bias scan-matching correspondences. MirrorDrift optimizes mirror placement, alignment, and actuation. In simulation, it increases the average pose error (APE) by 6.1x over random placement, degrading three SLAM systems to 2.29-3.31 m mean APE. In real-world experiments on a modern LiDAR with state-of-the-art interference mitigation, it induces localization errors of up to 6.03 m. To the best of our knowledge, this is the first successful SLAM-targeted attack against production-grade secure LiDARs. |
|
| Sapling-NeRF: Geo-Localised Sapling Reconstruction in Forests for Ecological Monitoring | 2026-02-26 | ShowSaplings are key indicators of forest regeneration and overall forest health. However, their fine-scale architectural traits are difficult to capture with existing 3D sensing methods, which make quantitative evaluation difficult. Terrestrial Laser Scanners (TLS), Mobile Laser Scanners (MLS), or traditional photogrammetry approaches poorly reconstruct thin branches, dense foliage, and lack the scale consistency needed for long-term monitoring. Implicit 3D reconstruction methods such as Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) are promising alternatives, but cannot recover the true scale of a scene and lack any means to be accurately geo-localised. In this paper, we present a pipeline which fuses NeRF, LiDAR SLAM, and GNSS to enable repeatable, geo-localised ecological monitoring of saplings. Our system proposes a three-level representation: (i) coarse Earth-frame localisation using GNSS, (ii) LiDAR-based SLAM for centimetre-accurate localisation and reconstruction, and (iii) NeRF-derived object-centric dense reconstruction of individual saplings. This approach enables repeatable quantitative evaluation and long-term monitoring of sapling traits. Our experiments in forest plots in Wytham Woods (Oxford, UK) and Evo (Finland) show that stem height, branching patterns, and leaf-to-wood ratios can be captured with increased accuracy as compared to TLS. We demonstrate that accurate stem skeletons and leaf distributions can be measured for saplings with heights between 0.5m and 2m in situ, giving ecologists access to richer structural and quantitative data for analysing forest dynamics. |
|
| Multi-robot LiDAR SLAM: a practical case study in underground tunnel environments | 2026-02-25 | ShowMulti-robot SLAM aims at localizing and building a map with multiple robots, interacting with each other. In the work described in this article, we analyze the pipeline of a decentralized LiDAR SLAM system to study the current limitations of the state of the art, and we discover a significant source of failures, i.e., that the loop detection is the source of too many false positives. We therefore develop and propose a new heuristic to overcome these limitations. The environment taken as reference in this work is the highly challenging case of underground tunnels. We also highlight potential new research areas still under-explored. |
14 pages, 14 figures |
| Title | Date | Abstract | Comment |
|---|---|---|---|
| Towards Path-Creative Navigation: Robot Navigation through Embodied Interaction | 2026-10-09 | ShowAutonomous navigation in cluttered and constrained environments typically assumes a fixed environment and searches only for paths within existing free space. However, a route can be blocked by an articulated structure, a movable object, or a pedestrian, and reaching the goal may therefore require appropriate embodied interaction with the environment. This paper formulates Path-Creative Navigation (PCN), a navigation paradigm in which the robot recovers free space by coordinating locomotion and embodied interaction, and addresses one class of PCN tasks with a mapless framework. The vision-LiDAR-odometry framework integrates local observations, obstacle geometry, and robot-state estimates through a unified navigation system, providing stable semantic and geometric information for navigation and interaction. The traversability-aware decision method utilizes visual and LiDAR distance information to determine whether a blockage is actionable and whether it can be safely bypassed, enabling the robot to interact only when necessary while supporting articulated structure pushing, movable object pushing, obstacle avoidance, and pedestrian requests. We design simulation scenarios and evaluate different methods in both simulation and two real-world tasks that cannot be completed by conventional navigation alone without embodied interaction. These results demonstrate that our framework achieves higher task completion than the baseline methods by coordinating navigation and interaction in cluttered and constrained environments. |
|
| Degeneracy-Orthogonal Geometric Constraints for LiDAR SLAM | 2026-09-29 | ShowAutonomous robot navigation relies on simultaneous localization and mapping (SLAM) to estimate motion and maintain an accurate pose within an environment. However, in axially uniform corridors such as long tunnels and pipelines, LiDAR odometry is fundamentally limited by unconstrained drift along the feature-weak travel direction. This structural degeneracy cannot be resolved by local scan matching alone. To address this challenge, we propose the Degeneracy-orthogonal Contour Offset Descriptor (DeCOD), a structure-aligned geometric descriptor for cross-sectional landmarks. Cross-sectional boundaries, such as pipe joints and structural rings, provide metric constraints along this degenerate axis, but distinguishing individual landmarks requires capturing subtle surface variations across nearly identical profiles. The descriptor parameterizes signed normal deviation from estimated boundary contours, and matching explicitly resolves heading ambiguity and decouples first-order contour errors by distortion estimation. Matched landmarks yield geometric factors that enforce agreement in cross-section position and corridor axis alignment during pose-graph optimization, correcting longitudinal drift while leaving rotation about the common axis unconstrained. On a public benchmark and in field experiments, DeCOD achieves robust landmark retrieval over standard 3D descriptors and successfully stabilizes trajectories across different odometry frontends, reliably constraining longitudinal drift under geometric degeneracy. |
8 pages, 4 figures |
| ArborSplat: Online Semantic Gaussian Splatting SLAM for Orchards | 2026-09-22 | ShowOrchard robots need maps that preserve small but semantically important structures such as trunks, trellises, and fruit. 3D Gaussian Splatting (3DGS) SLAM achieves high photometric fidelity. However, its optimization remains appearance-driven, and transferring image semantics to 3D points is unreliable for thin structures, whose pixels may receive depth from background surfaces. We present ArborSplat, an online semantic 3DGS SLAM system that tracks with LiDAR odometry and optimizes semantics directly on the Gaussian map, constrained by class-specific height bands above a ground plane fitted to each keyframe's stereo point cloud, and fuses multi-view evidence into a semantic point cloud online while rejecting labels inconsistent with the local ground surface or with monocular depth. Class-constrained refinement reserves Gaussian capacity for underrepresented structures and, under reduced budgets, increases training-view accuracy on tree classes. We evaluate the approach on apple and pear orchards during dormancy, flowering, and harvesting. On full routes, it keeps ATE below 0.5 m on all 12 traversals. On shared 301-frame segments, it exceeds SGS-SLAM and GS3LAM by 0.23 to 0.50 training-view and 0.15 to 0.36 held-out mIoU while running 1.7 to 7.5 times faster, whereas SemGauss-SLAM runs out of GPU memory on all six. |
8 pag...8 pages, 4 figures, 4 tables |
| TRaIL-Odom: Tightly Coupled Continuous Time Radar-IMU-LiDAR Odometry with Adaptive Doppler Weighting | 2026-09-03 | ShowExisting radar-LiDAR fusion methods rely on fixed residual weights, even though the informativeness of radar Doppler and LiDAR geometry is scan- and direction-dependent, leading to uniform radar weighting that misallocates Doppler information across translational directions. To address this limitation, we propose two degeneracy-aware Doppler reweighting modules within a tightly coupled Radar-IMU-LiDAR odometry framework: per-point radar reweighting and scan-wise radar gain scheduling. Since geometric degeneracy is directional, we first identify weak translational directions from the LiDAR geometry and reweight individual radar Doppler constraints based on their alignment with the weak subspace. We further adjust the overall radar contribution using LiDAR geometric anisotropy such that radar is emphasized when LiDAR observability is poor and suppressed when LiDAR constraints are already reliable. Across 13 evaluated sequences, TRaIL-Odom achieves state-of-the-art overall performance, with clear advantages in geometrically degenerate scenes. In ablation experiments on three degenerate sequences, combining the two adaptive weighting modules reduces RMSE ATE and RTE by 86.0% and 78.5% relative to the fixed-weight baseline. We make our code and an accompanying dataset publicly available at https://github.com/ChiyunNoh/TRaIL-Odom. |
Accep...Accepted for publication at the IEEE Robotics and Automation Letters on 23 August, 2026 |
| LF-GICP: Parameter-Free Degeneracy-Aware LiDAR Odometry via a Voxel-Normal Localizability Field | 2026-08-20 | ShowScan-to-map LiDAR odometry drifts unboundedly along the unobservable axes of geometrically degenerate environments like tunnels and corridors, and existing degeneracy handling requires environment-specific parameter tuning. This paper presents a parameter-free approach. We show that in voxelized GICP the Gauss--Newton (GN) Hessian masks translational degeneracy, because covariance regularization keeps the translation block artificially well-conditioned. We bypass this with a regularization-free voxel-normal localizability field and two of its statistics: a normalized fraction |
|
| Marker-Constrained Pose-Graph Correction for Cross-Platform Georeferencing in GNSS-Denied Environments | 2026-08-17 | ShowAutonomous operation in GNSS-denied environments requires heterogeneous mapping pipelines to maintain a consistent spatial reference. This paper presents a framework using camouflage-matched fiducial markers fabricated from Cholesteric Spherical Reflectors (CSRs) as pre-surveyed visual anchors. The anchors georeference both a lightweight LiDAR-odometry trajectory and a dense RTAB-Map reconstruction, allowing their outputs to be expressed in a common LUREF frame (geodetic coordinate reference system used in Luxembourg) without requiring GNSS measurements during operation. The method combines coarse similarity alignment with marker-constrained pose-graph optimization. We evaluate it using two handheld acquisition sessions with ground-level and elevated motion profiles emulating UGV and UAV operation. A single iMarker was relocated among six surveyed positions, with the first position revisited to quantify drift correction. Marker-anchor correction reduced revisit inconsistency by 97.9% and 99.1% for the UAV- and UGV-emulating sessions, respectively, and improved held-out anchor prediction compared with one-time alignment. Separately georeferenced dense reconstructions achieved a median cross-session nearest-neighbour distance of 58 cm without explicit cross-session registration. Marker processing operated in real time, while trajectory correction required less than 0.25 s per session. These results demonstrate a proof of concept for georeferencing lightweight odometry and dense reconstructions using visually unobtrusive, pre-surveyed anchors during GNSS-denied operation. |
14 pa...14 pages, 5 figures, 5 tables, submitted to SPIE Security + Defence conference |
| Online Lidar-Only Odometry with Retrospective Refinement of Overlapping Submaps | 2026-07-27 | ShowLidar odometry aims to estimate the ego-motion of a mobile platform from a stream of lidar scans. Traditional scan-to-map approaches register each scan against a single, evolving map, which propagates registration errors over time. To mitigate this, we propose a multi-submap approach where the current scan is registered against multiple overlapping submaps instead of a single static map. By optimizing the resulting constraints in a pose graph, our method enables not only precise estimation of the current pose, but also retrospective refinement of the submaps' anchor points, which improves short-term consistency and long-term accuracy. We demonstrate that our approach achieves significant accuracy gains over several state-of-the art lidar odometry methods on a variety of automotive datasets while simultaneously maintaining real-time performance. Ablation studies confirm the critical role of multiple registrations and retrospective refinement of the map as core factors for our accuracy gains. Code and raw results are available on our public GitHub at https://github.com/Fusion-Goettingen/RAL_2026_Kurda_Online. |
8 pag...8 pages, 3 figures, 5 tables |
| FAR-LIO: Enabling High-Speed Autonomy through Fast, Accurate, and Robust LiDAR-Inertial Odometry | 2026-06-24 | ShowRobust and accurate odometry estimation is essential in modern robotics. In environments characterized by highly dynamic motion and sensor noise, odometry estimation becomes increasingly challenging. Autonomous racing combines both factors in an unstructured setting, where minimizing odometry latency is essential for stable closed-loop control. This paper introduces FAR-LIO, a highly optimized CUDA-accelerated LiDAR-inertial odometry framework developed for Fast, Accurate, and Robust performance. Our system leverages a novel CUDA-based voxel hashmap to enable parallelized nearest-neighbor search and efficient map updates. We employ a sparsity-aware Generalized Iterative Closest Point algorithm with adaptive thresholding on top of the CUDA-based voxel hashmap with adaptive density to achieve low-latency without compromising accuracy. An Extended Kalman Filter serves as a robust backend. It utilizes an upsampling and delay compensation strategy to fuse the LiDAR odometry with high-frequency IMU data, thereby ensuring a robust and smooth odometry output. We evaluate FAR-LIO across four different sensor setups, using both public datasets and data from two autonomous racecars driving at speeds of up to 250 km/h. FAR-LIO achieves an average 6.9% reduction in the positional error and 38.4% lower runtime compared to state-of-the-art baselines on target hardware using a single parameter set. This demonstrates its computational efficiency and broad applicability. To build upon our work, our code is available open-source on https://github.com/TUMFTM/FAR-LIO. |
8 pag...8 pages, accepted for publication at IROS2026 (IEEE/RSJ International Conference on Intelligent Robots and Systems 2026) |
| Motor Angular Speed Preintegration for Multirotor UAV State Estimation | 2026-06-18 | ShowA precise state estimate is crucial for a tight feedback control that enables agile and near-obstacle flights of UAVs. The state-of-the-art methods fuse slow pose measurements with high-frequency inertial measurements to obtain a precise state estimate. However, the inertial measurements from the IMU onboard the UAV are degraded by vibrations from spinning propellers and the precision of the estimated state suffers. We propose a novel approach based on the preintegration of accelerations obtained from motor speeds. We show that the accelerations obtained in this manner can be used for state propagation on their own to achieve better precision without including the IMU. Further, we propose a factor composed of the preintegrated motor speeds that can be directly employed in factor graph optimization frameworks. We combine our factor with LiDAR measurements into the proposed Motor Angular Speed LiDAR Odometry (MAS-LO) algorithm for precise state estimation, which we open-source. Lastly, we evaluate the estimation precision against a state-of-the-art inertial algorithm LIO-SAM to show 28% improvement in position and 65% in velocity estimation accuracy, 14% lower measurement lag, and high robustness to wrong parameter values. |
|
| Semantic-weighted ICP for LiDAR Odometry: Class-Aware Residual Reweighting for Robust Scan Registration | 2026-06-02 | ShowLiDAR odometry is a fundamental component of autonomous robotic systems, relying on geometric registration between consecutive point clouds to estimate ego-motion. However, traditional geometric approaches often degrade in dynamic or unstructured environments due to unreliable correspondences caused by moving objects, sparse geometric features, vegetation, and semantically ambiguous structures. Existing works have shown that, some of these limitations can be addressed by introducing semantic information from the environment in the registration process. In this work, we build on this, and show that not all elements in the environment are equally relevant for registration. Hence, we propose a semantic class-weighted ICP for LiDAR odometry. Instead of strictly filtering out points belonging to specific semantic classes, the proposed approach weights the residuals of points belonging to semantic categories based on their expected geometric stability. This strategy enables informative but potentially unstable structures, to contribute to the registration process while mitigating the influence of dynamic objects. The experimental evaluation was conducted on the SemanticKITTI and RELLIS-3D datasets, which include urban, highway, rural, and off-road environments. The empirical results show that the proposed Semantic-weighted ICP improves pose estimation, especially in challenging off-road scenarios where conventional rigid features are scarce. Furthermore, the analysis reveals that the effectiveness of this weighting strategy is highly environment-dependent, influenced by the structural and semantic composition of the scene. |
|
| SOCC-ICP: Semantics-Assisted Odometry based on Occupancy Grids and ICP | 2026-05-14 | ShowReliable pose estimation in previously unseen environments is a fundamental capability of autonomous systems. Existing LiDAR odometry methods typically employ point-, surfel-, or NDT-based map representations, which are distinct from the semantic occupancy grids commonly used for downstream tasks such as motion planning. We introduce SOCC-ICP, a semantics-assisted odometry framework that jointly performs Semantic OCCupancy grid mapping and LiDAR scan alignment. Each map voxel encodes geometric and semantic statistics, enabling adaptive point-to-point or point-to-plane ICP based on local planarity. Further, the occupancy grid naturally filters dynamic objects through raycasting-based free-space updates. Across diverse evaluation scenarios, SOCC-ICP achieves performance competitive with state-of-the-art LiDAR odometry and remains robust in geometrically degenerate environments, even in the absence of semantic cues. When semantic labels are available, integrating them into map construction, downsampling, and correspondence weighting yields further accuracy gains. By unifying odometry and semantic occupancy grid mapping within a single representation, SOCC-ICP eliminates redundant map structures and directly provides a map suitable for downstream robotic applications. |
9 pag...9 pages, 3 figures, Accepted May 2026 for publication in IEEE Robotics and Automation Letters (RA-L) |
| OptMap: Geometric Map Distillation via Submodular Maximization | 2026-05-12 | ShowAutonomous robots rely on geometric maps to inform a diverse set of perception and decision-making algorithms. As autonomy requires reasoning and planning on multiple scales, each algorithm may require a different map for optimal performance. LiDAR sensors generate an abundance of geometric data (up to 50 MB per second) to satisfy these diverse requirements. However, the point-based operations required to process perception data are both memory and computationally expensive. Such operations can be bypassed via learned representations that encode similarity, but selecting informative, size-constrained maps remains an NP-hard combinatorial problem. In this work we present OptMap: a geometric map distillation algorithm which achieves online, application-specific map generation via multiple theoretical and algorithmic innovations. A central feature is the maximization of set functions that exhibit diminishing returns, i.e., submodularity, using polynomial-time algorithms with provably near-optimal solutions. We formulate a novel submodular reward function which quantifies informativeness, reduces input set sizes, and minimizes solution bias. Further, we propose a dynamically reordered streaming submodular algorithm which improves empirical solution quality and addresses input order bias via an online approximation of the value of all scans. Testing was conducted on open-source and custom datasets with an emphasis on long-duration mapping sessions, highlighting OptMap's minimal computation requirements. OptMap's practical value is then illustrated through its application to online geometric change detection. Open-source ROS1 and ROS2 packages are available and can be used alongside any LiDAR odometry algorithm. |
|
| Hi-LOAM: Hierarchical Implicit Neural Fields for LiDAR Odometry and Mapping | 2026-04-02 | ShowLiDAR Odometry and Mapping (LOAM) is a pivotal technique for embodied-AI applications such as autonomous driving and robot navigation. Most existing LOAM frameworks are either contingent on the supervision signal, or lack of the reconstruction fidelity, which are deficient in depicting details of large-scale complex scenes. To overcome these limitations, we propose a multi-scale implicit neural localization and mapping framework using LiDAR sensor, called Hi-LOAM. Hi-LOAM receives LiDAR point cloud as the input data modality, learns and stores hierarchical latent features in multiple levels of hash tables based on an octree structure, then these multi-scale latent features are decoded into signed distance value through shallow Multilayer Perceptrons (MLPs) in the mapping procedure. For pose estimation procedure, we rely on a correspondence-free, scan-to-implicit matching paradigm to estimate optimal pose and register current scan into the submap. The entire training process is conducted in a self-supervised manner, which waives the model pre-training and manifests its generalizability when applied to diverse environments. Extensive experiments on multiple real-world and synthetic datasets demonstrate the superior performance, in terms of the effectiveness and generalization capabilities, of our Hi-LOAM compared to existing state-of-the-art methods. |
This ...This manuscript is the accepted version of IEEE Transactions on Multimedia |
| Omni-LIVO: Robust RGB-Colored Multi-Camera Visual-Inertial-LiDAR Odometry via Photometric Migration and ESIKF Fusion | 2026-03-29 | ShowWide field-of-view (FoV) LiDAR sensors provide dense geometry across large environments, but existing LiDAR-inertial-visual odometry (LIVO) systems generally rely on a single camera, limiting their ability to fully exploit LiDAR-derived depth for photometric alignment and scene colorization. We present Omni-LIVO, a tightly coupled multi-camera LIVO system that leverages multi-view observations to comprehensively utilize LiDAR geometric information across extended spatial regions. Omni-LIVO introduces a Cross-View direct alignment strategy that maintains photometric consistency across non-overlapping views, and extends the Error-State Iterated Kalman Filter (ESIKF) with multi-view updates and adaptive covariance. The system is evaluated on public benchmarks and our custom dataset, showing improved accuracy and robustness over state-of-the-art LIVO, LIO, and visual-inertial SLAM baselines. Code and dataset will be released upon publication. |
Accep...Accepted by IEEE Robotics and Automation Letters (RA-L). Early Access version available. This version supersedes all previous versions and is the official accepted manuscript for citation |
| Evaluating and Improving the Robustness of LiDAR Odometry and Localization Under Real-World Corruptions | 2026-02-22 | ShowLiDAR odometry and localization are two widely used and fundamental applications in robotic and autonomous driving systems. Although state-of-the-art (SOTA) systems achieve high accuracy on clean point clouds, their robustness to corrupted data remains largely unexplored. We present the first comprehensive benchmark to evaluate the robustness of LiDAR pose-estimation techniques under 18 realistic synthetic corruptions. Our results show that, under these corruptions, odometry position errors escalate from 0.5% to more than 80%, while localization performance stays consistently high. To address this sensitivity, we propose two complementary strategies. First, we design a lightweight detection-and-filter pipeline that classifies the point cloud corruption and applies a corresponding filter (e.g., bilateral filter for noise) to restore the point cloud quality. Our classifier accurately identifies each corruption type, and the filter effectively restores odometry accuracy to near-clean data levels. Second, for learning-based systems, we show that fine-tuning using the corrupted data substantially improves robustness across all tested corruptions and even boosts performance on clean point clouds on one data sequence. |
|
| R-VoxelMap: Accurate Voxel Mapping with Recursive Plane Fitting for Online LiDAR Odometry | 2026-01-18 | ShowThis paper proposes R-VoxelMap, a novel voxel mapping method that constructs accurate voxel maps using a geometry-driven recursive plane fitting strategy to enhance the localization accuracy of online LiDAR odometry. VoxelMap and its variants typically fit and check planes using all points in a voxel, which may lead to plane parameter deviation caused by outliers, over segmentation of large planes, and incorrect merging across different physical planes. To address these issues, R-VoxelMap utilizes a geometry-driven recursive construction strategy based on an outlier detect-and-reuse pipeline. Specifically, for each voxel, accurate planes are first fitted while separating outliers using random sample consensus (RANSAC). The remaining outliers are then propagated to deeper octree levels for recursive processing, ensuring a detailed representation of the environment. In addition, a point distribution-based validity check algorithm is devised to prevent erroneous plane merging. Extensive experiments on diverse open-source LiDAR(-inertial) simultaneous localization and mapping (SLAM) datasets validate that our method achieves higher accuracy than other state-of-the-art approaches, with comparable efficiency and memory usage. Co |