意图修订下的承诺层级：工具使用智能体中打捞行为的信念修订解释
Commitment Hierarchies under Intent Revision: A Belief-Revision Account of Salvage in Tool-Use Agents
浏览论文内容
中文总结 AI 辅助
本研究提出通过一次性修订分类与确定性决策传播，使工具使用智能体在用户中途改变意图时实现成本最优的打捞，比重启便宜43%且保持100%正确率。
中文摘要 AI 辅助
当用户在任务中途改变主意时，已经将任务拆分为子目标并为工具调用付费的智能体必须决定，对于每个缓存的子结果，是保留、修补还是丢弃它（打捞）；重新开始会浪费有效工作，而继续不变则回答的是旧问题。我们的主要发现是，打捞质量是角色设计问题而非模型能力问题：一个语言模型被逐个节点询问保留/修补/丢弃问题时不可靠，但被要求一次性对修订进行分类，并通过确定性层传播决策时，它在来自两家供应商的所有三个测试模型上达到了成本最优的预言机。将计划建模为承诺层级，并将意图变化建模为具有AGM风格公设的信念修订算子，我们证明了仅观察节点局部视图的策略不可能同时安全且成本最优，而单一分类设计则两者兼得。在三个环境中，该策略恢复了全部可实现的节省，比重新开始便宜43%，且正确率为100%。
英文摘要
When a user changes their mind partway through a task, an agent that has already split the task into sub-goals and paid for tool calls must decide, per cached sub-result, whether to keep, patch, or discard it (salvage), restarting wastes valid work and continuing unchanged answers the old question. Our main finding is that salvage quality is a matter of role design rather than model capability: a language model asked the keep/patch/discard question one node at a time is unreliable, but asked to classify the revision once, with a deterministic layer propagating the decision, it reaches the cost-optimal oracle on all three models tested, from two vendors. Modeling the plan as a commitment hierarchy and the intent change as a belief-revision operator with AGM style postulates, we prove that no policy observing only a node's local view can be both safe and cost optimal, while the single classification design is both. Across three environments the policy recovers the full achievable savings, 43% cheaper than restart, at 100% correctness.
发表机构
- Georgia Institute of Technology（佐治亚理工学院）
机构由 AI 辅助整理，请以论文原文为准。
