Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
[NeurIPS 2025] Official implementation of "RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics"
| Date | Stars |
|---|---|
| 2026-07-24 | 263 |
| 2026-07-25 | 263 |
| 2026-07-28 | 265 |
| 2026-07-30 | 265 |
| 2026-08-06 | 265 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
<h1 align="center">RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics</h1>
<h3 align="center">From words to exactly where you mean — with RoboRefer </h3>
<p align="center">
<a href="https://arxiv.org/abs/2506.04308"><img src="https://img.shields.io/badge/arXiv-2506.04308-b31b1b.svg?logo=arxiv" alt="arXiv"></a>
<a href="https://zhoues.github.io/RoboRefer/"><img src="https://img.shields.io/badge/%F0%9F%8F%A0%20Project-Homepage-blue" alt="Project Homepage"></a>
<a href="https://huggingface.co/datasets/JingkunAn/RefSpatial"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Dataset-RefSpatial--Dataset-brightgreen" alt="Dataset"></a>
<a href="https://huggingface.co/datasets/BAAI/RefSpatial-Bench"><img src="https://img.shields.io/badge/🤗%20Benchmark-RefSpatial--Bench-green.svg" alt="Benchmark"></a>
<a href="https://huggingface.co/collections/Zhoues/roborefer-and-refspatial-6857c97848fab02271310b89"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Weights-RoboRefer-yellow" alt="Weights"></a>
</p>
<div style="text-align: center; background-color: white;">
<img src="assets/motivation.png" width=100% >
</div>
## 🔥 Updates
[2025-12-16] 🔥🔥🔥 We’ve released RoboTracer, a more powerful version of RoboRefer, on [arxiv](https://arxiv.org/abs/2512.13660) and launched the [project page](https://zhoues.github.io/RoboTracer/). It retains all RoboRefer features while adding multi-step, metric-grounded spatial tracing with explicit reasoning. Come and try the newer dataset TraceSpatial and the more challenging benchmark TraceSpatial-Bench!
[2025-10-25] 🔥🔥🔥 [RefSpatial-Expand-Bench](https://huggingface.co/datasets/JingkunAn/RefSpatial-Expand-Bench) is now released! Based on the original benchmark, the new version expands indoor scenes (e.g., factories, stores) and introduces previously **unexplored outdoor scenarios (e.g., streets, parking lots)**, offering a **more comprehensive evaluation of spatial referring tasks**.
[2025-09-26] 🔥🔥🔥 RefSpatial-Bench is now used by [Qwen3-VL](https://github.com/QwenLM/Qwen3-VL) and [Gemini-Robotics-1.5](https://deepmind.google/discover/blog/gemini-robotics-15-brings-ai-agents-into-the-physical-world/) to evaluate spatial reasoning in complex embodied scenarios! Come on, test your model with RefSpatial-Bench!
<table>
<tr>
<th>Qwen3-VL Results</th>
<th>Gemini-Robotics-1.5 Results</th>
</tr>
<tr>
<td><img width="1749" height="747" alt="Qwen-3-VL" src="https://github.com/user-attachments/assets/852d42fb-76b4-40b7-81ba-577040d62fac" /></td>
<td><img width="1420" height="787" alt="Gemini-Robotics-1 5-2" src="https://github.com/user-attachments/assets/f7529979-36e0-4d83-946c-882b4ac69e69" /></td>
</table>
[2025-09-18] 🔥🔥🔥 RoboRefer gets accepted to NeurIPS 2025! See you in Mexico City and San Diego, USA!
[2025-07-29] We release the SFT-trained 8B model and test examples of outdoor scenes.
[2025-07-01] We release the RefSpatial Dataset and SFT training code.
[2025-06-23] We release the SFT-trained 2B model and inference code with RefSpatial-Bench evaluation code.
[2025-06-06] RefSpatial-Bench is released on [HF](https://huggingface.co/datasets/BAAI/RefSpatial-Bench). Let's evaluate your model's spatial referring ability!
[2025-06-06] RoboRefer is released on [arxiv](https://arxiv.org/abs/2506.04308) and the project page is set up at [here](https://zhoues.github.io/RoboRefer/).
## 🤗 Model Zoo & Dataset & Benchmark
<table>
<tr>
<th>Model/Dataset/Benchmark</th>
<th>Note</th>
</tr>
<tr>
<td><a href="https://huggingface.co/Zhoues/NVILA-2B-Depth">NVILA-2B-Depth</a></td>
<td> The base model with depth encoder initialized from the image encoder. </td>
</tr>
<tr>
<td><a href="https://huggingface.co/Zhoues/RoboRefer-2B-Depth-Align">RoboRefer-2B-Align</a></td>
<td> The 1st SFT step of the 2B model for depth alignment. </td>
</tr>
<tr>
<td><a hrExcerpt of 21,822 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:fa192cb6254a1beb, topic:vision-language-model, desc:vision-language, readme:vision-language
matched fp:fa192cb6254a1beb, topic:robotics, desc:robotics, readme:robotics