研究人员提出环境演化方法，让终端智能体训练环境持续变难，在Qwen模型上提升了18%性能。
该研究提出环境演化方法，通过离线增加环境难度，在训练过程中持续提供学习信号。研究团队从多轮学习目标中推导出三种影响环境难度的演化方向，并通过循环工程多智能体工具实现。在Hy4预览版、Claude Opus 5和GPT-5.6 Sol的定量滚动实验中，环境演化方法 consistently 生成更具挑战性的环境。在Qwen3.6-27B和Qwen3.6-35B-A3B的简单长程强化学习训练中，该方法使它们在Terminal-Bench 2.1上的性能分别提升了14.4和18.0个百分点。
Environment Evolution for Terminal Agents
Scaling interactive and verifiable environments is critical for training terminal agents. As frontier models become more capable, environments synthesized from scratch become less challenging and thus provide limited learning signals. Recent co-evolution methods iteratively synthesize environments near the model's learnable frontier based on weaknesses exposed during rollouts. However, their dependence on on-policy rollouts limits generalization and the continuous provision of learning signals as the model becomes stronger. In this paper, we propose environment evolution, which incrementally increases environment difficulty off-policy and schedules the evolved environments generation by generation during training to provide continuous learning signals. We derive three evolution directions that influence environment difficulty from the multi-turn learning objective and then implement evolution along these directions through a loop-engineered multi-agent harness. Quantitative rollout experiments with Hy4 preview, Claude Opus 5, and GPT-5.6 Sol show that environment evolution consistently produces more difficult environments. We validate its effectiveness on Qwen3.6-27B and Qwen3.6-35B-A3B through simple long-horizon RL training, improving their performance by 14.4 and 18.0 percentage points on Terminal-Bench 2.1, respectively.