QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents

Published 16 October 2026

QwenGyre is a reinforcement learning framework for training language-model agents on tasks lasting hours and using hundreds of interactions. It dynamically shifts GPUs between rollout and training, while reducing repeated paths in complex trajectories. Tests show faster training and a 6-point improvement on NL2RepoBench.