arXiv cs.CL· Tiejin Chen, Xiaoou Liu, Michael Clifford, Ahmadreza Moradipari, Yitao Chen, Kyungtae Han, Nejib Ammar, Hua Wei·· 11 小时前AI 评分31
先检测再消歧：LLM 基于扰动保留的输入歧义检测方法
Detect Before You Disambiguate: Perturbation-Based Input Ambiguity Detection in LLMs
AI 导读
研究提出一种标签保持的扰动框架，用于在消歧之前检测 LLM 输入歧义，包含黑盒规则扰动库和白盒重要性引导两种检测器。在开放域、对话和视觉问答四个数据集上，该方法优于输出侧不确定性基线和基于 LLM 改写的扰动基线。论文已被 AACL-IJCNLP 2026 主会接收，代码已开源。
正文 · 原文
Abstract:Large language models (LLMs) often answer ambiguous inputs as if they had a single intended meaning. This is problematic because input ambiguity is an upstream source of uncertainty: when a prompt admits multiple valid interpretations, the key question is whether the model should answer directly at all. Existing input uncertainty methods often rely on LLM-generated statements or clarifications, but such transformations may already disambiguate the input before the ambiguity is measured. We propose a label-preserving perturbation framework for input ambiguity detection in which perturbations remain close to the original input while preserving its label. We instantiate this principle with two detectors: a black-box rule-based perturbation library and a white-box importance-guided method using model-internal signals. Experiments across four datasets in open-domain, conversational, and visual QA show that our methods outperform output-side uncertainty baselines and LLM-paraphrase-based perturbation baselines. Our results highlight the importance of detecting ambiguity before attempting to disambiguate it. Our code is available \href{this https URL}{\textcolor{purple}{here}}.
| Comments: | Accepted to AACL-IJCNLP 2026 Main |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.36353 [cs.CL] |
| (or arXiv:2610.36353v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.36353 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tiejin Chen [view email]
[v1]
Mon, 26 Oct 2026 22:18:10 UTC (382 KB)
来源：arXiv cs.CL · arxiv.org