发表机构
Beijing Normal University; Ke Holdings（北京师范大学; 贝壳找房）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对LLM智能体恢复仅按平均成功率评估的缺陷，提出因果干预路由器（CIR），在ALFWorld任务上将成功率从70.33%提升至73.33%，并区分挽救与干扰效应。
AI 中文摘要
大型语言模型智能体依赖外部脚手架（harness）在模型与环境之间传递信息，并从执行错误中恢复。然而，恢复通常仅通过平均任务成功率来评判。这掩盖了一个重要的矛盾：同一操作既能挽救失败的轨迹，也可能干扰原本会成功的轨迹。我们将恢复问题构建为因果决策问题。从相同的执行状态出发，我们比较有恢复和无恢复两种情况的结果，区分“挽救”与“干扰”，并研究恢复的价值随时间如何变化。随后，我们引入因果干预路由器（Causal Intervention Router, CIR），这是一种轻量级策略，利用恢复前可用的信息来决定何时值得进行干预。在Qwen3-14B的长时程ALFWorld任务中，CIR将成功率从70.33%提升至73.33%，提高了3.00个百分点。它不改变所有评估轨迹中具有正确观测的部分。额外对照实验表明，恢复的收益不能仅由环境返回的新观测所解释。这些结果提供了一种实用的方法来评估恢复并选择性地应用它。
英文摘要
Large language model agents rely on external harnesses to pass information between the model and its environment and to recover from execution errors. Yet recovery is usually judged only by average task success. This hides an important tension. The same operation can rescue a failing trajectory or disrupt one that would otherwise succeed. We frame recovery as a causal decision problem. Starting from the same execution state, we compare what happens with and without recovery, separate rescue from harm, and study how the value of recovery changes over time. We then introduce the Causal Intervention Router (CIR), a lightweight policy that uses information available before recovery to decide when intervention is worthwhile. On long-horizon ALFWorld tasks with Qwen3-14B, CIR raises success from 70.33% to 73.33%, a gain of 3.00 percentage points. It leaves all evaluated trajectories with correct observations untouched. Additional controls show that the benefit of recovery cannot be explained solely by the new observation returned by the environment. These results provide a practical way to evaluate recovery and apply it selectively.
