arXiv cs.AI· Xinghao Li, Chenchen Tan, Youyang Qu, Shujie Cui, Xingliang Yuan, Cunjian Chen, Chi Liu, Longxiang Gao·· 11 小时前AI 评分32
多模态大推理模型中的推理感知跨模态关联遗忘
Reasoning-Aware Cross-Modal Association Unlearning in Multimodal Large Reasoning Models
AI 导读
研究提出 Visual-Keyed Counterfactual Rebinding（VKCR），一种参数高效的跨模态关联编辑框架，用于在多模态大推理模型（MLRMs）中抑制指定视觉上下文下的目标知识行为，同时保留该知识在允许上下文中的访问。
正文 · 原文
Abstract:Multimodal large reasoning models (MLRMs) integrate visual inputs with textual and parametric knowledge, allowing sensitive information to be accessed through cross-modal associations during reasoning. This raises privacy concerns when a visual input should no longer provide access to such information. Machine unlearning can remove this influence without retraining the model from scratch. However, multimodal unlearning methods mainly target samples, concepts, or knowledge content, without separating the knowledge from the cross-modal association through which it is accessed. In reasoning models, target information may disappear from the final answer while remaining in intermediate reasoning, or be replaced by incorrect substitute knowledge. We therefore study cross-modal association unlearning, where the forget target is a designated visual-to-knowledge association: the model should suppress target-bearing behavior under the designated visual context while preserving access to the same knowledge in permitted contexts. We propose Visual-Keyed Counterfactual Rebinding (VKCR), a parameter-efficient framework for cross-modal association editing. VKCR first localizes the edit to a compact association-conditioned parameter subspace. It then reduces support for the designated target at task outputs and counterfactually rebinds the affected reasoning behavior by discouraging target-bearing or unsupported continuations while favoring valid non-sensitive alternatives. We further introduce CLEAR-R and MLLMU-R to evaluate target leakage, hallucinated reassociation, reasoning preservation, and modality-specific utility. Experiments show that VKCR substantially reduces image-conditioned target leakage and hallucinated reassociation, including across heterogeneous task formulations, while preserving non-sensitive visual reasoning and permitted text-only knowledge.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.06168 [cs.CV] |
| (or arXiv:2611.06168v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2611.06168 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Chenchen Tan [view email]
[v1]
Wed, 4 Nov 2026 08:48:07 UTC (318 KB)
来源：arXiv cs.AI · arxiv.org