行动前先选择：面向长程工具使用智能体的比较价值估计

AI 摘要

论文提出 CITA 方法，训练比较推理模型（CIM）来估计长程工具调用序列中候选下一步工具调用的长期价值，从而在执行前做出更优选择。CIM 利用真实轨迹配对、贝叶斯工具图模拟器和 LLM 语义比较三种信号进行训练。在三个工具使用基准和多个骨干 LLM 上，CITA 一致提升了 Tool F1 和任务成功率。

正文节选

Choosing Before Acting: Comparative Value Estimation for Long-Horizon Tool-Use Agents Abstract Large language models (LLMs) rely on long-horizon tool invocation sequences for complex tasks, where each invocation can alter the task state and condition subsequent decisions. In long-horizon tool use, final-outcome rewards provide weak credit assignment over long interaction traces. Step-level rewards can offer more targeted feedback, but obtaining reliable step supervision often requires human or L

发布时间：2026-10-06 12:00

抓取时间：2026-10-05 12:04

来源机构：arXiv