让大模型主动提问，解决任务条件不全时的推理错误问题。
Teaching LLMs to Ask: Self-Querying Category-Theoretic Planning for Under-Specified Reasoning
- 通过自查询和假设补全机制，动态处理未知前提条件。
- 在维基百科和食谱任务中，资源违规率分别降至14.9%和5.8%。
- 适合需要可靠推理、避免幻觉的复杂规划场景使用。
大语言模型在推理时规划常因信息不全而失效：当关键前提未明确时，模型易虚构事实或生成违反硬约束的计划。本文提出自查询双向范畴规划（SQ-BCP），显式表示前提状态（满足/违反/未知），并通过向外部源提问或引入附加动作建立缺失前提来解决未知。该方法采用双向搜索，并使用基于拉回的验证器作为目标兼容性的范畴证明，仅用距离评分进行排序与剪枝。理论上保证：当验证器成功且硬约束通过确定性检查时，接受的计划与目标要求兼容；在分支因子有限、分辨率深度受限条件下，若存在可接受计划，SQ-BCP 必能发现。在维基如何（WikiHow）和食谱生成（RecipeNLG）任务中，针对隐藏前提，其资源违规率分别降至14.9%和5.8%（优于最优基线的26.0%和15.7%），同时保持良好参考质量。
原文摘要 · Abstract (English)
Inference-time planning with large language models frequently breaks under partial observability: when task-critical preconditions are not specified at query time, models tend to hallucinate missing facts or produce plans that violate hard constraints. We introduce \textbf{Self-Querying Bidirectional Categorical Planning (SQ-BCP)}, which explicitly represents precondition status (\texttt{Sat}/\texttt{Viol}/\texttt{Unk}) and resolves unknowns via (i) targeted self-queries to an oracle/user or (ii) \emph{bridging} hypotheses that establish the missing condition through an additional action. SQ-BCP performs bidirectional search and invokes a pullback-based verifier as a categorical certificate of goal compatibility, while using distance-based scores only for ranking and pruning. We prove that when the verifier succeeds and hard constraints pass deterministic checks, accepted plans are compatible with goal requirements; under bounded branching and finite resolution depth, SQ-BCP finds an accepting plan when one exists. Across WikiHow and RecipeNLG tasks with withheld preconditions, SQ-BCP reduces resource-violation rates to \textbf{14.9\%} and \textbf{5.8\%} (vs.\ \textbf{26.0\%} and \textbf{15.7\%} for the best baseline), while maintaining competitive reference quality.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。