何时提问何时作答：对话中歧义指代消解的策略性澄清研究
When to Ask, When to Answer: Strategic Clarification for Ambiguous Reference Resolution in Dialogue
AI 导读
一项新研究探讨大语言模型在歧义指代消解中的“策略性澄清”问题，即何时提问、何时直接作答。研究者基于 SIMMC 2.1 数据集构建受控测试环境，记录消解每个歧义指代所需的提问次数，并比较基线 LLM 与常见后训练方法的表现。
正文 · 原文
Abstract:For humans, referential disambiguation is a collaborative process involving clarification questions asked when they will meaningfully change the outcome. However, large language models (LLMs) have been shown to be biased toward premature answering, preferring definite single-turn responses over more clarification exchanges. In this paper, we study the problem of "strategic clarification" -- asking clarification questions in a reference game when their answers will narrow the set of candidate referents. We construct a controlled test environment from the SIMMC 2.1 dataset that records how many questions each ambiguous reference needs to be successfully resolved, and examine how baseline LLM behavior and common post-training methods ask clarifying questions vs. attempt to answer the referent. We also propose BACE (Belief-Aware Critic with Epistemic tracking), a lightweight action chooser method on top of a frozen LM backbone, and show that BACE more consistently chooses the optimal number of clarifying questions than methods such as GRPO and ArCHer.
| Comments: | 22 pages, 10 figures. Code and data: this https URL |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.05517 [cs.CL] |
| (or arXiv:2611.05517v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.05517 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ibrahim Khebour [view email]
[v1]
Tue, 3 Nov 2026 19:14:42 UTC (1,503 KB)
来源：arXiv cs.AI · arxiv.org