Data as of Sep 18, 2026 · Based on 3,315,446 AI responses across 10,525 prompts · See how Parse measures this
AgentGym-RL is a modular framework for training autonomous LLM agents to perform long-horizon, multi-turn decision-making through reinforcement learning without relying on supervised fine-tuning. It features a decoupled architecture (environment, agent, training module) and supports diverse real-world scenarios—web navigation, deep search, digital games, embodied tasks, and scientific tasks—along with mainstream RL algorithms, plus the ScalingInter-RL method to balance exploration and exploitation over longer horizons. The project reports that agents match or surpass commercial models on 27 tasks and plans to open-source the complete AgentGym-RL framework, including code and datasets, developed by Fudan University, ByteDance Seed, and Shanghai Innovation Institute.
Parse Score
Excerpts where AgentGym-RL appeared in the AI's answer
AgentGym-RL / ScalingInter-RL : An open-source framework designed specifically for training LLM agents via RL in multi-turn, long-horizon interactive environments.
AgentGym-RL : A purpose-built open framework specifically engineered to train Large Language Model (LLM) agents for long-horizon decision-making
