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.
verl-agent is an extension of veRL, designed for training LLM/VLM agents via RL. verl-agent is also the official code for paper "Group-in-Group Policy Optimization for LLM Agent Training"
| Date | Stars |
|---|---|
| 2026-07-24 | 2151 |
| 2026-07-25 | 2153 |
| 2026-07-28 | 2153 |
| 2026-07-30 | 2153 |
| 2026-07-31 | 2165 |
| 2026-08-06 | 2165 |
Today
— stars today
This week
+12 stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.56%/day
<p align="center">
<img src="./docs/gigpo/logo-verl-agent.png" alt="logo" width="55%">
</p>
<h3 align="center">
<b>Group-in-Group Policy Optimization for LLM Agent Training</b>
<br>
<b>NeurIPS 2025</b>
</h3>
<p align="center">
<a href="https://arxiv.org/abs/2505.10978">
<img src="https://img.shields.io/badge/arXiv-Paper-red?style=flat-square&logo=arxiv" alt="arXiv Paper"></a>
<a href="https://github.com/langfengQ/verl-agent">
<img src="https://img.shields.io/badge/GitHub-Project-181717?style=flat-square&logo=github" alt="GitHub Project"></a>
<a href="https://huggingface.co/collections/langfeng01/verl-agent-684970e8f51babe2a6d98554">
<img src="https://img.shields.io/badge/HuggingFace-Models-yellow?style=flat-square&logo=huggingface" alt="HuggingFace Models"></a>
<a href="https://x.com/langfengq/status/1930848580505620677">
<img src="https://img.shields.io/badge/Twitter-Channel-000000?style=flat-square&logo=x" alt="X Channel"></a>
<a href="https://github.com/langfengQ/verl-agent/blob/master/LICENSE">
<img src="https://img.shields.io/badge/license-Apache%202.0-blue.svg?style=flat-square" alt="License"></a>
<a href="https://github.com/langfengQ/verl-agent/issues">
<img src="https://img.shields.io/github/issues/langfengQ/verl-agent?style=flat-square&color=green" alt="GitHub issues"></a>
<a href="https://github.com/langfengQ/verl-agent/stargazers">
<img src="https://img.shields.io/github/stars/langfengQ/verl-agent?style=social" alt="Repo stars"></a>
</p>
`verl-agent` is an extension of [veRL](https://github.com/volcengine/verl), specifically designed for training **large language model (LLM) agents via reinforcement learning (RL)**.
Unlike prior approaches that simply concatenate full interaction histories, `verl-agent` proposes **step-independent multi-turn rollout mechanism**, which allows for **fully customizable** per-step input structures, history management, and memory modules. This design makes `verl-agent` **highly scalable for very long-horizon, multi-turn RL training** (e.g., tasks in ALFWorld can require up to 50 steps to complete).
`verl-agent` provides a **diverse set of RL algorithms** (including our new algorithm GiGPO) and a **rich suite of agent environments**, enabling the development of reasoning agents in both visual and text-based tasks.
# News
- [2026.05] `GraphGPO` accepted at [ICML 2026](https://icml.cc/)! 🎉🎉🎉 [[Paper](https://arxiv.org/abs/2605.26684)] [[Code](https://github.com/langfengQ/verl-agent/tree/master/recipe/GraphGPO)]
- [2026.02] `HGPO` accepted at [ICLR 2026](https://iclr.cc/)! 🎉🎉🎉 [[Paper](https://openreview.net/forum?id=T8Dev99qnz)] [[Code](https://github.com/langfengQ/verl-agent/tree/master/recipe/hgpo)]
- [2026.02] 🔥 We open-source [Dr. MAS](https://github.com/langfengQ/DrMAS), which supports stable end-to-end RL post-training of **multi-agent LLM systems**! [[Paper](https://arxiv.org/pdf/2602.08847)] [[Code](https://github.com/langfengQ/DrMAS)]
- [2025.12] `Qwen3-VL` is supported! See example [here](./examples/gigpo_trainer/run_sokoban_qwen3vl.sh).
- [2025.09] `GiGPO` is now supported by [ROLL](https://github.com/alibaba/ROLL)! [[Document](https://alibaba.github.io/ROLL/docs/English/UserGuide/agentic/agentic_GiGPO)] [[Train Curves](https://github.com/alibaba/ROLL/issues/173#issuecomment-3332106534)].
- [2025.09] `verl-agent`-style training pipeline is now supported by [OpenManus-RL](https://github.com/OpenManus/OpenManus-RL)!
- [2025.09] [GiGPO](https://arxiv.org/abs/2505.10978) accepted at [NeurIPS 2025](https://neurips.cc/)! 🎉🎉🎉
- [2025.08] Add **Search-R1 experiments** and **similarity-based GiGPO**! Check out GiGPO's superior performance in Search-R1 experiments [here](#results).
- [2025.07] `GiGPO` & `verl-agent` talks at [Agent for SWE meetup](https://lu.ma/e498qhsi) by LF AI & Data Singapore on 7/11.
- [2025.07] Add modular memory manager. See [here](./agent_system/memExcerpt of 30,620 characters
Read on GitHub
121
haibin · Bytedance Seed
65
63
32
Yaowei Zheng · Millennium Science School · China
8
Joel · @bytedance · China
8
8
Shawn/Yuxuan Tong · ByteDance Seed · China
7
7
Willem Jiang
6
6
Lumeng Wu
5
Xingyao Wang · OpenHands / All Hands AI
5
4
4
Ikko Eltociear Ashimine · Japan
3
Ze-Yi LIN · Emotion Machine Lab · China
3
fzyzcjy · +=1 (seriously this is the name)
3
zhou fan · China
3
Franz Srambical · p(doom) · Germany
3
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:c342763a70dff7a9, topic:agent-framework, readme:multi-agent, readme:multi agent
matched fp:c342763a70dff7a9, topic:large-language-models
matched fp:c342763a70dff7a9, topic:reinforcement-learning, readme:reinforcement learning, readme:rl algorithms