AI Paper Atlas

Research map / Large language models / Agents, code and math

MedAgentGym: A Scalable Agentic Training Environment for Code-Centric Reasoning in Biomedical Data Science

Ran Xu, Yuchen Zhuang, Yishan Zhong, Yue Yu, Zifeng Wang, Xiangru Tang, Hang Wu, May Dongmei Wang, Peifeng Ruan, Donghan Yang, Tao Wang, Guanghua Xiao and 4 more

ICLR 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents · 43 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add SFT on successful gpt-4.1-mini trajectories (ablation)Qwen2.5-7B-Instruct · MedAgentGym avgaverage score: 16.89 -> 53.87Improved
add DPO after SFT (ablation)Qwen2.5-7B-Instruct · MedAgentGym avgaverage score: 53.87 -> 59.90Improved
use GRPO instead of DPO after SFT (ablation)Qwen2.5-7B-Instruct · MedAgentGym avgaverage score: 59.90 -> 62.17Improved
use GRPO instead of PPO after SFT (ablation)Qwen2.5-14B-Instruct · MedAgentGym avgaverage score: 69.56 -> 71.42Improved
use GRPO instead of PPO after SFT (ablation)Qwen2.5-7B-Instruct · MedAgentGym avgaverage score: 57.96 -> 62.17Improved
train with SFT or DPO on in-distribution tasks, evaluate OOD (ablation)Qwen2.5-7B-Instruct · external OOD averageaverage score: 20.60 -> 20.91 (SFT), 20.84 (DPO)No clear effect
train with GRPO, evaluate OOD (ablation)Qwen2.5-7B-Instruct · external OOD averageaverage score: 20.60 -> 36.63Improved
train with DPO, evaluate OOD (ablation)Qwen2.5-14B-Instruct · external OOD averageaverage score: 27.92 -> 27.91No clear effect
remove debugging capability from interactive coding (ablation)Med-Copilot · MedAgentGym taskssuccess rateImproved
increase sampled trajectories from K=1 to K=16 (ablation)Med-Copilot verifier · MedAgentGymPass@K: 17.0 -> 45.0Improved
add iterative DPO rounds after rejection-sampling SFT (ablation)Qwen2.5-7B-Instruct · MedAgentGymsuccess rate: diminishing returnsMixed
use coding-specialized or medical-reasoning LLMs instead of general LLMsLLM zero-shot · MedAgentGym avgaverage score: HuatuoGPT-o1-7B 16.11 vs Qwen2.5-7B-Instruct 17.43Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy