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.
[ICML 2026] World-R1: Reinforcing 3D Constraints for Text-to-Video Generation
| Date | Stars |
|---|---|
| 2026-07-24 | 410 |
| 2026-07-25 | 409 |
| 2026-07-28 | 411 |
| 2026-07-30 | 411 |
| 2026-07-31 | 413 |
| 2026-08-06 | 413 |
Today
— stars today
This week
+2 stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.49%/day
<p align="center"> <img src="https://microsoft.github.io/World-R1/assets/banner.png" alt="World-R1" width="480"/> </p> <h1 align="center">Reinforcing 3D Constraints for Text-to-Video Generation</h1> <p align="center"> <a href="https://arxiv.org/abs/2604.24764"><img src="https://img.shields.io/badge/Paper-B31B1B?style=for-the-badge&logo=arxiv&logoColor=white" alt="Paper"></a> <a href="https://aka.ms/world-r1"><img src="https://img.shields.io/badge/Project%20Page-000000?style=for-the-badge&logo=googlechrome&logoColor=white" alt="Project Page"></a> <a href="https://huggingface.co/datasets/microsoft/World-R1"><img src="https://img.shields.io/badge/Dataset-FFD21E?style=for-the-badge&logo=huggingface&logoColor=black" alt="Dataset"></a> </p> <p align="center"> <a href="https://lhmd.top/">Weijie Wang</a><sup>1,2,*†</sup> <a href="https://github.com/Shredded-Pork">Xiaoxuan He</a><sup>1,*</sup> <a href="https://github.com/Tacossp">Youping Gu</a><sup>1,*</sup> <a href="https://www.microsoft.com/en-us/research/people/yifanyang/">Yifan Yang</a><sup>2,‡</sup> <a href="https://steve-zeyu-zhang.github.io/">Zeyu Zhang</a><sup>3</sup> <a href="https://openreview.net/profile?id=~Yefei_He1">Yefei He</a><sup>1</sup> <a href="https://github.com/DINGYANB">Yanbo Ding</a><sup>2</sup> <a href="https://openreview.net/profile?id=~Xirui_Hu1">Xirui Hu</a><sup>3</sup> <a href="https://donydchen.github.io/">Donny Y. Chen</a><sup>3</sup> <a href="https://www.microsoft.com/en-us/research/people/zhiyuhe/">Zhiyuan He</a><sup>2</sup> <a href="https://www.microsoft.com/en-us/research/people/yuqyang/">Yuqing Yang</a><sup>2,‡</sup> <a href="https://bohanzhuang.github.io/">Bohan Zhuang</a><sup>1,‡</sup> </p> <p align="center"> <sup>1</sup>Zhejiang University <sup>2</sup>Microsoft Research <sup>3</sup>Independent Researcher </p> <p align="center"> <img src="https://microsoft.github.io/World-R1/assets/teaser.jpg" alt="World-R1 teaser" width="100%"> </p> World-R1 aligns text-to-video generation with 3D constraints through reinforcement learning. Instead of changing the base video model architecture or relying on large-scale 3D supervision, it combines camera-aware latent initialization, 3D-aware rewards from pre-trained foundation models, and a periodic decoupled training strategy to improve geometric consistency while preserving visual quality and motion diversity. ## Highlights - 3D-aware reinforcement learning aligns generated videos with geometric constraints through meta-view assessment, reconstruction consistency, and trajectory alignment rewards. - General visual quality is preserved by combining the 3D-aware reward with an aesthetic reward during Flow-GRPO-based post-training. - A periodic dynamic-only training phase regularizes the model with dynamic-scene prompts, improving motion diversity while retaining learned 3D consistency. - Camera-aware latent initialization converts text-specified camera motion into trajectory-guided noise wrapping, enabling implicit camera conditioning without changing the base video architecture. ## Method <p align="center"> <img src="https://microsoft.github.io/World-R1/assets/pipeline.jpg" alt="World-R1 pipeline" width="100%"> </p> World-R1 first converts camera instructions in text prompts into explicit trajectories and injects the motion prior into the initial video latents through noise wrapping. During RL fine-tuning, the model is optimized with 3D-aware feedback from reconstruction and camera-control metrics, together with a general visual reward. A periodic dynamic-only phase prevents the model from overfitting to rigid static scenes. ## Setup Use a Python 3.10+ environment with CUDA and a PyTorch build that matches your driver. A practical setup flow is: 1. Create and activate a clean environment: ```bash conda create -n world-r1 python=3.10 -y conda activa
Excerpt of 10,301 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:c524dc79c4e0e0d0, topic:video-generation, desc:video generation, readme:video generation
matched fp:c524dc79c4e0e0d0, topic:reinforcement-learning, readme:reinforcement learning