LEMUR: Latent Entropy-aware Multimodal Unlearning via Visual-anchored Reasoning Redirection

• URL: http://arxiv.org/abs/2608.11691v1
• Date: Wed, 12 Aug 2026 06:03:50 GMT
• Title: LEMUR: Latent Entropy-aware Multimodal Unlearning via Visual-anchored Reasoning Redirection
• Abstract summary: Reinforcement-learning (RL) post-training equips multimodal large reasoning models (MLRMs) with exploratory chains of thought (CoT)<n>We find that even when a sensitive fact is successfully unlearned from the final answer, the model may still reproduce it in its reasoning trace.<n>We propose LEMUR, a fully training-free, inference-time unlearning framework for natively RL-trained multimodal models.
• Score: 12.083572608650904
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reinforcement-learning (RL) post-training equips multimodal large reasoning models (MLRMs) with exploratory chains of thought (CoT), substantially improving visual reasoning. However, we find that this capability introduces a distinct privacy vulnerability: even when a sensitive fact is successfully unlearned from the final answer, the model may still reproduce it in its reasoning trace. This leakage is substantially more pronounced in natively RL-trained MLRMs than in their non -reasoning base models, revealing a privacy risk that existing unlearning methods are not designed to address. We show that RL-induced exploration leaves sensitive content with a distinctive token-level entropy signature that is largely absent from base models. Based on this observation, we propose LEMUR, a fully training-free, inference-time unlearning framework for natively RL-trained multimodal models. LEMUR uses entropy dynamics as a control signal to identify when sensitive reasoning begins and when sanitization should stop. During this interval, it redirects the reasoning trajectory through entropy-modulated visual-anchor latent injection, replacing committed tokens with sanitized, probability-weighted embeddings re-grounded in the input image. Across diverse MLRMs, LEMUR consistently outperforms existing unlearning met hods in suppressing both reasoning-trace and answer leakage, while better preserving non-sensitive utility and output fluency. These results demonstrate that RL-induced entropy dynamics provide a distinctive signal for privacy leakage and that exploiting this signal enables effective training-free unlearning for reasoning-capable multimodal models.

Related papers

• TTSD-FAR: Test-Time Self-Distillation with Fisher-Anchored Restoration for Missing-Modality Emotion Recognition in LVLMs [83.78708832901194]
  Large video-language models (LVLMs) have shown remarkable performance on multimodal tasks like multimodal emotion recognition (ER) in the wild.<n>However, real-world deployment remains challenging: modalities may be missing or noisy at test time.<n>We propose a solution in which a frozen teacher guides an adaptive low-rank student via self-distillation.<n>Experiments on MELD, DFEW, and BAH under 0%-50% missing modalities show that this unified adaptation-restoration design consistently outperforms entropy-based adaptation, RAG, and perplexity-based generation.
  arXiv  Detail & Related papers  (2026-08-18T23:40:28Z)
• Balancing Efficiency and Efficacy: Training-Free Attention-Guided Switching Between Explicit and Latent Thoughts for MLLMs [44.74237674161132]
  Reasoning in Multimodal Large Language Models (MLLMs) requires both fine-grained visual perception and rigorous logical deduction.<n>Explicit text-based Chain-of-Thought (CoT) is computationally expensive and prone to visual hallucinations.<n>Existing latent reasoning methods typically require costly training.
  arXiv  Detail & Related papers  (2026-08-04T10:46:10Z)
• Uncertainty-Aware Exploratory Direct Preference Optimization for Multimodal Large Language Models [53.15468578562038]
  We propose an Uncertainty-aware Exploratory Direct Preference Optimization (UE-DPO) method for Multimodal Large Language Models (MLLMs)<n>We first quantify the uncertainty from the model's failure to ground token predictions in the given image.<n>We then encourage more learning pressure on visually deficient tokens in preferred samples, and alleviate the over-penalization of beneficial knowledge in dispreferred samples.
  arXiv  Detail & Related papers  (2026-05-06T13:08:12Z)
• Reflect to Inform: Boosting Multimodal Reasoning via Information-Gain-Driven Verification [55.357038267439684]
  Visual Re-Examination (VRE) is a self-evolving training framework that enables MLLMs to autonomously perform visual introspection during reasoning without additional visual inputs.<n>VRE consistently improves reasoning accuracy and perceptual reliability, while substantially reducing hallucinations, especially in long-chain settings.
  arXiv  Detail & Related papers  (2026-03-27T12:22:13Z)
• ReLAPSe: Reinforcement-Learning-trained Adversarial Prompt Search for Erased concepts in unlearned diffusion models [12.021923446217722]
  Machine unlearning is a key defense mechanism for removing unauthorized concepts from text-to-image diffusion models.<n>Existing adversarial approaches for exploiting this leakage are constrained by fundamental limitations.<n>We introduce ReLAPSe, a policy-based adversarial framework that reformulates concept restoration as a reinforcement learning problem.
  arXiv  Detail & Related papers  (2026-01-30T21:56:50Z)
• MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory [46.632646462295234]
  We propose MemRL, a framework that enables agents to self-evolve via non-parametric reinforcement learning on episodic memory.<n>MemRL employs a Two-Phase Retrieval mechanism that filters candidates by semantic relevance and then selects them based on learned Q-values.<n>Our analysis experiments confirm that MemRL effectively reconciles the stability-plasticity dilemma, enabling continuous runtime improvement without weight updates.
  arXiv  Detail & Related papers  (2026-01-06T17:14:50Z)
• Towards Reasoning-Preserving Unlearning in Multimodal Large Language Models [17.184948937224142]
  Machine unlearning aims to erase requested data from trained models without full retraining.<n> intermediate chain-of-thought steps can still leak sensitive information even when final answers are forgotten.<n>We propose R-MUSE, a training-free and inference-time intervention framework that steers internal representations to forget both answers and reasoning traces.
  arXiv  Detail & Related papers  (2025-11-26T13:45:52Z)
• Reasoning Model Unlearning: Forgetting Traces, Not Just Answers, While Preserving Reasoning Skills [42.1825027925353]
  Large reasoning models (LRMs) have enabled strong chain-of-thought (CoT) generation through test-time computation.<n>We show that conventional unlearning algorithms, originally designed for non-reasoning models, are inadequate for LRMs.<n>We propose Reasoning-aware Representation Misdirection for Unlearning ($R2MU$), a novel method that effectively suppresses sensitive reasoning traces.
  arXiv  Detail & Related papers  (2025-06-15T20:54:23Z)
• Hybrid Latent Reasoning via Reinforcement Learning [50.6763762323985]
  We explore latent reasoning by leveraging the capabilities of large language models (LLMs) via reinforcement learning (RL)<n>We introduce hybrid reasoning policy optimization (HRPO), an RL-based hybrid latent reasoning approach that integrates prior hidden states into sampled tokens with a learnable gating mechanism.<n>HRPO-trained LLMs remain interpretable and exhibit intriguing behaviors like cross-lingual patterns and shorter completion lengths.
  arXiv  Detail & Related papers  (2025-05-24T01:26:16Z)
• Unconditional Truthfulness: Learning Unconditional Uncertainty of Large Language Models [104.55763564037831]
  We train a regression model that leverages attention maps, probabilities on the current generation step, and recurrently computed uncertainty scores from previously generated tokens.<n>Our evaluation shows that the proposed method is highly effective for selective generation, achieving substantial improvements over rivaling unsupervised and supervised approaches.
  arXiv  Detail & Related papers  (2024-08-20T09:42:26Z)
• Analyzing Adversarial Inputs in Deep Reinforcement Learning [53.3760591018817]
  We present a comprehensive analysis of the characterization of adversarial inputs, through the lens of formal verification. We introduce a novel metric, the Adversarial Rate, to classify models based on their susceptibility to such perturbations. Our analysis empirically demonstrates how adversarial inputs can affect the safety of a given DRL system with respect to such perturbations.
  arXiv  Detail & Related papers  (2024-02-07T21:58:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.