{"id":28366962,"url":"https://github.com/thuml/rlvr-world","last_synced_at":"2025-08-29T09:07:19.358Z","repository":{"id":294488914,"uuid":"985088554","full_name":"thuml/RLVR-World","owner":"thuml","description":"Official repository for \"RLVR-World: Training World Models with Reinforcement Learning\", https://arxiv.org/abs/2505.13934","archived":false,"fork":false,"pushed_at":"2025-06-09T12:11:33.000Z","size":14128,"stargazers_count":45,"open_issues_count":0,"forks_count":2,"subscribers_count":2,"default_branch":"main","last_synced_at":"2025-06-20T20:50:53.315Z","etag":null,"topics":["grpo","real2sim","reinforcement-learning-with-verifiable-rewards","rlvr","robotic-manipulation","text-game","verl","video-generation","video-gpt","video-prediction","web-agent","world-model"],"latest_commit_sha":null,"homepage":"https://thuml.github.io/RLVR-World/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/thuml.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-05-17T03:39:30.000Z","updated_at":"2025-06-17T15:46:30.000Z","dependencies_parsed_at":"2025-06-09T13:33:59.067Z","dependency_job_id":null,"html_url":"https://github.com/thuml/RLVR-World","commit_stats":null,"previous_names":["thuml/rlvr-world"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/thuml/RLVR-World","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/thuml%2FRLVR-World","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/thuml%2FRLVR-World/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/thuml%2FRLVR-World/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/thuml%2FRLVR-World/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/thuml","download_url":"https://codeload.github.com/thuml/RLVR-World/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/thuml%2FRLVR-World/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":272659047,"owners_count":24971604,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-08-29T02:00:10.610Z","response_time":87,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["grpo","real2sim","reinforcement-learning-with-verifiable-rewards","rlvr","robotic-manipulation","text-game","verl","video-generation","video-gpt","video-prediction","web-agent","world-model"],"created_at":"2025-05-29T00:13:41.814Z","updated_at":"2025-08-29T09:07:19.350Z","avatar_url":"https://github.com/thuml.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# RLVR-World: Training World Models with Reinforcement Learning\n\n[![Project Page](https://img.shields.io/badge/Project_Page-blue)](https://thuml.github.io/RLVR-World/)\n[![Paper](https://img.shields.io/badge/arXiv-Paper-b31b1b.svg?logo=arxiv)](https://arxiv.org/abs/2505.13934)\n[![Hugging Face](https://img.shields.io/badge/Hugging_Face-Models_\u0026_Datasets-F8D44E.svg?logo=huggingface)](https://huggingface.co/collections/thuml/rlvr-world-682f331c75a904b8febc366a)\n\nThis is the official code base for the paper [RLVR-World: Training World Models with Reinforcement Learning](https://arxiv.org/abs/2505.13934).\n\nGive it a star 🌟 if you find our work useful!\n\n## 🔥 News\n\n- 🚩 **2025.05.26**: We release all models and datasets.\n- 🚩 **2025.05.21**: We open-source our training codes.\n- 🚩 **2025.05.21**: Our paper is released on [arXiv](https://arxiv.org/abs/2505.13934).\n\n## 📋 TL;DR\n\nWe pioneer training world models through RLVR:\n\n- World models across various modalities (particularly, language and videos) are unified under a sequence modeling formulation;\n- Task-specific prediction metrics serve as verifiable rewards directly optimized by RL.\n\n![concept](assets/concept.png)\n\n## 🤗 Models and Datasets\n\nAt the moment, we provide the following models and datasets:\n\n| Modality | Type        | Domain             | Name                                                         |\n| -------- | ----------- | ------------------ | ------------------------------------------------------------ |\n| Language | Dataset     | Text game          | [bytesized32-world-model-cot](https://huggingface.co/datasets/thuml/bytesized32-world-model-cot) |\n| Language | World model | Text game          | [bytesized32-world-model-sft](https://huggingface.co/thuml/bytesized32-world-model-sft) |\n| Language | World model | Text game          | [bytesized32-world-model-rlvr-binary-reward](https://huggingface.co/thuml/bytesized32-world-model-rlvr-binary-reward) |\n| Language | World model | Text game          | [bytesized32-world-model-rlvr-task-specific-reward](https://huggingface.co/thuml/bytesized32-world-model-rlvr-task-specific-reward) |\n| Language | Dataset     | Web navigation     | [webarena-world-model-cot](https://huggingface.co/datasets/thuml/webarena-world-model-cot) |\n| Language | World model | Web navigation     | [webarena-world-model-sft](https://huggingface.co/thuml/webarena-world-model-sft) |\n| Language | World model | Web navigation     | [webarena-world-model-rlvr](https://huggingface.co/thuml/webarena-world-model-rlvr) |\n| Video    | Tokenizer   | Robot manipulation | [rt1-frame-tokenizer](https://huggingface.co/thuml/rt1-frame-tokenizer) |\n| Video    | World model | Robot manipulation | [rt1-world-model-single-step-base](https://huggingface.co/thuml/rt1-world-model-single-step-base) |\n| Video    | World model | Robot manipulation | [rt1-world-model-single-step-rlvr](https://huggingface.co/thuml/rt1-world-model-single-step-rlvr) |\n| Video    | Tokenizer   | Robot manipulation | [rt1-compressive-tokenizer](https://huggingface.co/thuml/rt1-compressive-tokenizer) |\n| Video    | World model | Robot manipulation | [rt1-world-model-multi-step-base](https://huggingface.co/thuml/rt1-world-model-multi-step-base) |\n| Video    | World model | Robot manipulation | [rt1-world-model-multi-step-rlvr](https://huggingface.co/thuml/rt1-world-model-multi-step-rlvr) |\n\n## 💬 Evaluating Language World Models\n\nSee [`lang_wm`](/lang_wm):\n\n- Text game state prediction\n- Web page state prediction\n- Application: Model predictive control for web agents\n\n## 🎇 Evaluating Video World Models\n\nSee [`vid_wm`](/vid_wm):\n\n- Robot manipulation trajectory prediction\n- Application: Real2sim policy evaluation\n\n## 🎥 Showcases\n\n![showcase](assets/showcase.png)\n\n## 🚀 Release Progress\n\n- [x] Video world model with RLVR\n- [x] Pre-trained \u0026 post-trained video world model weights\n- [x] Real2sim policy evaluation with video world models\n- [x] Text game SFT data\n- [x] Web page SFT data\n- [x] Language world model on text games with RLVR\n- [x] Language world model on web pages with RLVR\n- [x] Post-trained language world model weights\n- [x] Web agents with language world models\n\n## 📜 Citation\n\nIf you find this project useful, please cite our paper as:\n\n```\n@article{wu2025rlvr,\n    title={RLVR-World: Training World Models with Reinforcement Learning}, \n    author={Jialong Wu and Shaofeng Yin and Ningya Feng and Mingsheng Long},\n    journal={arXiv preprint arXiv:2505.13934},\n    year={2025},\n}\n```\n\n## 🤝 Contact\n\nIf you have any questions, please contact wujialong0229@gmail.com.\n\n## 💡 Acknowledgement\n\nWe sincerely appreciate the following github repos for their valuable codebase we build upon:\n\n- https://github.com/volcengine/verl\n- https://github.com/thuml/iVideoGPT\n- https://github.com/kyle8581/WMA-Agents\n- https://github.com/cognitiveailab/GPT-simulator\n- https://github.com/web-arena-x/webarena\n- https://github.com/simpler-env/SimplerEnv\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fthuml%2Frlvr-world","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fthuml%2Frlvr-world","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fthuml%2Frlvr-world/lists"}