Tag
提出 Harness Scaling 这一新的强化学习扩展维度，通过在多样化执行配置上训练以提升智能体模型对未见 harness 的泛化能力，并在多个基准上验证其优于单一 harness 训练的效果。
StateM is a runtime system that enhances long-horizon agent execution through durable states and recoverable runbooks, achieving 95.3% accuracy on Terminal-Bench 2.1 and significantly reducing API costs.