Alibaba Qwen Unveils QwenGyre for Extreme-Long-Horizon AI Agents
Alibaba's Qwen team has introduced QwenGyre, an innovative elastic reinforcement learning framework designed to train coding agents over extreme-long-horizon tasks. Capable of handling rollouts lasting hours and processing nearly one million tokens using unmodified black-box harnesses, this system represents a significant step forward in developing advanced, highly autonomous AI agents capable of complex problem-solving.
Source: MarkTechPost