WooooDyy/AgentGym-RL
A reinforcement learning framework for training LLM-based agents to solve complex, multi-turn decision-making tasks across diverse real-world environments.
Not currently ranked — collecting fresh signals.
star history
AgentGym-RL provides a training framework and benchmark for developing autonomous LLM agents capable of sequential decision-making through multi-turn reinforcement learning interactions. It supports mainstream RL algorithms and evaluates agents across 27 tasks spanning diverse environments including web navigation, tool use, and code generation. The benchmark assesses open-source 7B-scale models against commercial baselines.
Frequently asked
- What is WooooDyy/AgentGym-RL?
- A reinforcement learning framework for training LLM-based agents to solve complex, multi-turn decision-making tasks across diverse real-world environments.
- Is AgentGym-RL open source?
- Yes — WooooDyy/AgentGym-RL is open source, released under the MIT license.
- What language is AgentGym-RL written in?
- WooooDyy/AgentGym-RL is primarily written in Python.
- How popular is AgentGym-RL?
- WooooDyy/AgentGym-RL has 818 stars on GitHub.
- Where can I find AgentGym-RL?
- WooooDyy/AgentGym-RL is on GitHub at https://github.com/WooooDyy/AgentGym-RL.