Alibaba Qwen introduces QwenGyre, an elastic RL framework for long-horizon coding agents
AIAlibaba's Qwen team introduced QwenGyre, an online reinforcement learning framework that trains coding agents through unmodified black-box harnesses on runs of hours and about 700K tokens. It moves idle GPUs between rollout and training and caps each execution at five trajectory samples.
Source details
MarkTechPost · Asif Razzaq · marktechpost.comPublished · added here
Original title: Alibaba Qwen Introduces QwenGyre: Elastic Reinforcement Learning RL for XLong-Horizon Agents