arXiv cs.AI· Gongli Xi, Kun Wang, Zeming Gao, Huahui Yi, Haolang Lu, Qianshan Wei, Ye Tian, Wendong Wang·· 5 小时前AI 评分44
ClueTracer：免训练的多模态推理幻觉抑制方法
ClueTracer: Question-to-Vision Clue Tracing for Training-Free Hallucination Suppression in Multimodal Reasoning
AI 导读
研究者提出 ClueTracer，一个免训练、无参数、架构无关的插件，用于抑制多模态推理模型的幻觉。该工作定义了视觉线索检索指标 ClueRecall，并从问题出发沿推理路径追踪关键线索，定位任务相关图像区域。
正文 · 原文
Abstract:Large multimodal reasoning models solve challenging visual problems via explicit long-chain inference: they gather visual clues from images and decode clues into textual tokens. Yet this capability also increases hallucinations, where the model generates content that is not supported by the input image or the question. To understand this failure mode, we identify \emph{reasoning drift}: during clue gathering, the model over-focuses on question-irrelevant entities, diluting focus on task-relevant cues and gradually decoupling the reasoning trace from visual grounding. As a consequence, many inference-time localization or intervention methods developed for non-reasoning models fail to pinpoint the true clues in reasoning settings. Motivated by these insights, we introduce ClueRecall, a metric for assessing visual clue retrieval, and present ClueTracer, a training-free, parameter-free, and architecture-agnostic plugin for hallucination suppression. ClueTracer starts from the question and traces how key clues propagate along the model's reasoning pathway (question $\rightarrow$ outputs $\rightarrow$ visual tokens), thereby localizing task-relevant patches while suppressing spurious attention to irrelevant regions. Remarkably, \textbf{without any additional training}, ClueTracer improves all \textbf{reasoning} architectures (including \texttt{R1-OneVision}, \texttt{Ocean-R1}, \texttt{MM-Eureka}, \emph{etc}.) by $\mathbf{1.21\times}$ on reasoning benchmarks
| Comments: | 27 pages, 12 figures, accept by EMNLP 2026 (Oral) |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2602.02004 [cs.CV] |
| (or arXiv:2602.02004v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2602.02004 arXiv-issued DOI via DataCite |
Submission history
From: Gongli Xi [view email]
[v1]
Mon, 2 Feb 2026 12:03:56 UTC (4,102 KB)
[v2]
Wed, 28 Oct 2026 09:01:36 UTC (9,533 KB)
来源：arXiv cs.AI · arxiv.org