SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning该项目在GitHub已获得847星标（65分叉），代码活跃度高。主要编程语言为Python，采用MIT开源协议。
项目简介
SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning
核心功能
- [05/10/2026] Released the code for SFT data generation under `examples/sft_data_generation/`. - [04/03/2026] Released the SFT dataset on [🤗HF](https://huggingface.co/datasets/Jianwen/SkillRL-SFT-Data)! - [03/02/2026] Due to an accidental misconfiguration, we lost several hundred GitHub stars. If you previously starred this repo, we'd appreciate a re-star ⭐! - [02/23/2026] We released all the model checkpoints on HuggingFace! Feel free to use them as warm starts for continued RL training. - [02/18/2026] The code of SkillRL was released! - [02/10/2026] SkillRL paper was released on [arXiv](https://arxiv.org/abs/2602.08234)!
使用指南
🚀 Getting Started
项目信息
| 项目 | 信息 |
|---|---|
| Star数 | 847 |
| Fork数 | 65 |
| 编程语言 | Python |
| 开源协议 | MIT |
| 创建时间 | 2026-02-08 |
| 最近更新 | 2026-05-17 |
| 主题标签 | 无 |
来源：[aiming-lab/SkillRL](https://github.com/aiming-lab/SkillRL)