斯坦福论文：智能体循环卡顿改 harness，计划质量差需训练权重
New Stanford paper finds that harness changes fix agents that loop or stall, while agents that deliver bad plans need weight training instead.
斯坦福新论文发现，智能体的失败可分为流程失败（如循环、步数预算耗尽）和内容失败（交付了糟糕的计划），前者可通过修改 harness（模型周边的提示词、工具与检查）修复，后者需要微调权重。在旅行规划基准上，由 LLM 驱动的循环重写 harness，最佳运行结果再用于微调模型。harness 演化把 Qwen3.5-4B 在留出任务上的成绩从 0.16 提升到 0.30，计划交付率从 55% 升至 90%；但 harness 修改从未减少糟糕计划的比例，而 LoRA 适配器把 Qwen3.5-9B 留出运行中的糟糕计划占比从 28% 降到 5%。
正文
当前显示原文，中文全文译文尚未就绪。
New Stanford paper finds that harness changes fix agents that loop or stall, while agents that deliver bad plans need weight training instead.
An agent can be improved by editing its harness, the prompts, tools, and checks around the model, or by fine-tuning its weights.
They sorted failed runs into process failures, such as loops and used-up step budgets, and content failures, where a poor plan was delivered. On a travel-planning benchmark, an LLM-driven loop rewrote the harness, and its best runs then fine-tuned the model.
Harness evolution lifted Qwen3.5-4B from 0.16 to 0.30 on held-out tasks, as plan delivery rose from 55% to 90%. Harness edits never shrank the share of poor plans, but a LoRA adapter cut them from 28% to 5% of Qwen3.5-9B's held-out runs.
Before improving an agent, label why its runs fail, then fix process failures in the harness and content failures in the weights.