AgentGym-RL: An Open-Source Framework to Train LLM Agents for Long-Horizon Decision Making via Multi-Turn RL

ICLROral2026

Authors: Zhiheng Xi, Jixuan Huang, Chenyang Liao, Baodai Huang, Jiaqi Liu, Honglin Guo, yajie yang, Rui Zheng, Junjie Ye, Jiazheng Zhang, Wenxiang Chen, Wei He, Yiwen Ding, Guanyu Li, Zehui Chen, Zhengyin Du, Xuesong Yao, Yufei Xu, Jiecao Chen, Tao Gui, Zuxuan Wu, Qi Zhang, Xuanjing Huang, Yu-Gang Jiang
Affiliation: Fudan University
Venue: ICLR 2026
Track: Oral

TL;DR

We present AgentGym-RL, a unified open-source framework for training LLM agents from scratch across diverse and realistic environments, and propose ScalingInter-RL, a staged training strategy for stable long-horizon RL training.

Opening excerpt from the authors’ abstract. source

Read the paper

OpenReview

Topics

agent llm

← All ICLR 2026 Oral papers · Browse the whole archive