Seeing Through the Chain: Mitigate Hallucination in Multimodal Reasoning Models via CoT Compression and Contrastive Preference Optimization

• URL: http://arxiv.org/abs/2602.03380v1
• Date: Tue, 03 Feb 2026 11:00:55 GMT
• Title: Seeing Through the Chain: Mitigate Hallucination in Multimodal Reasoning Models via CoT Compression and Contrastive Preference Optimization
• Abstract summary: multimodal reasoning models (MLRMs) remain prone to hallucinations, and effective solutions are still underexplored.<n>We propose C3PO, a training-based mitigation framework comprising textbfCompression and textbfPreference textbfOptimization.
• Score: 78.94590726578014
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While multimodal reasoning models (MLRMs) have exhibited impressive capabilities, they remain prone to hallucinations, and effective solutions are still underexplored. In this paper, we experimentally analyze the hallucination cause and propose C3PO, a training-based mitigation framework comprising \textbf{C}hain-of-Thought \textbf{C}ompression and \textbf{C}ontrastive \textbf{P}reference \textbf{O}ptimization. Firstly, we identify that introducing reasoning mechanisms exacerbates models' reliance on language priors while overlooking visual inputs, which can produce CoTs with reduced visual cues but redundant text tokens. To this end, we propose to selectively filter redundant thinking tokens for a more compact and signal-efficient CoT representation that preserves task-relevant information while suppressing noise. In addition, we observe that the quality of the reasoning trace largely determines whether hallucination emerges in subsequent responses. To leverage this insight, we introduce a reasoning-enhanced preference tuning scheme that constructs training pairs using high-quality AI feedback. We further design a multimodal hallucination-inducing mechanism that elicits models' inherent hallucination patterns via carefully crafted inducers, yielding informative negative signals for contrastive correction. We provide theoretical justification for the effectiveness and demonstrate consistent hallucination reduction across diverse MLRMs and benchmarks.

Related papers

• Reasoning Matters: Mitigate Hallucination in Multimodal Large Reasoning Models via Reasoning-Conditioned Preference Optimization [59.20570719781289]
  Multimodal Large Reasoning Models introduce the reasoning paradigm, demonstrating strong capabilities on complex vision-language tasks.<n>Existing training-based methods typically mitigate hallucinations through response-level direct preference optimization (DPO), where the Chain-of-Thought (CoT) and the final answer are treated as a monolithic output and optimized jointly.<n>We derive Reasoning-Conditioned Direct Preference Optimization (RC-DPO), which models the CoT as a condition for answer generation and contrasts the preference for the same preferred answer under different CoT conditions, promoting answer-supportive reasoning chain alignment.
  arXiv  Detail & Related papers  (2026-05-27T03:27:23Z)
• CHASD: Language Increment-Calibrated Contrastive Decoding against Hallucination in LVLMs [16.77211535169488]
  We propose Contrastive Hallucination-Aware Step-wise Decoding (CHASD) for Large Vision-Language Models.<n> CHASD uses an uncertainty-driven confidence gate to activate the contrastive branch only when the maximum probability of the next-token is less than the threshold.<n> Experiments show that CHASD improves hallucination-related metrics over strong training-free baselines with competitive inference efficiency.
  arXiv  Detail & Related papers  (2026-05-22T08:03:12Z)
• Cognitive Pivot Points and Visual Anchoring: Unveiling and Rectifying Hallucinations in Multimodal Reasoning Models [65.57985131861399]
  hallucinations are strongly correlated with cognitive bifurcation points that often exhibit high entropy states.<n>We propose V-STAR, a lightweight, holistic training paradigm designed to internalize visually aware reasoning capabilities.
  arXiv  Detail & Related papers  (2026-04-11T13:59:05Z)
• Focus Matters: Phase-Aware Suppression for Hallucination in Vision-Language Models [8.304027910542446]
  We investigate the internal attention dynamics of vision encoders in Large Vision-Language Models (LVLMs)<n>Our analysis reveals that hallucination behavior is particularly sensitive to tokens receiving low attention during the focus phase.<n>We propose a lightweight inference-time intervention that selectively suppresses such tokens during the focus phase.
  arXiv  Detail & Related papers  (2026-04-04T02:46:58Z)
• Thinking in Uncertainty: Mitigating Hallucinations in MLRMs with Latent Entropy-Aware Decoding [38.5840117402958]
  We argue that adequate contextual reasoning information can be directly extracted from the token probability distribution.<n>We present Latent Entropy-Aware Decoding, an efficient plug-and-play decoding strategy.<n>The model employs probability-weighted continuous embeddings under high-entropy states and transitions back to discrete token embeddings as entropy decreases.
  arXiv  Detail & Related papers  (2026-03-09T12:47:54Z)
• Imagination Helps Visual Reasoning, But Not Yet in Latent Space [65.80396132375571]
  We investigate the validity of latent reasoning using Causal Mediation Analysis.<n>We show that latent tokens encode limited visual information and exhibit high similarity.<n>We propose a straightforward alternative named CapImagine, which teaches the model to explicitly imagine using text.
  arXiv  Detail & Related papers  (2026-02-26T08:56:23Z)
• Analyzing Reasoning Consistency in Large Multimodal Models under Cross-Modal Conflicts [74.47786985522762]
  We identify a critical failure mode termed textual inertia, where models tend to blindly adhere to the erroneous text while neglecting conflicting visual evidence.<n>We propose the LogicGraph Perturbation Protocol that structurally injects perturbations into the reasoning chains of diverse LMMs.<n>Results reveal that models successfully self-correct in less than 10% of cases and predominantly succumb to blind textual error propagation.
  arXiv  Detail & Related papers  (2026-01-07T16:39:34Z)
• Look Closer! An Adversarial Parametric Editing Framework for Hallucination Mitigation in VLMs [6.645440928271175]
  Visionivate-Language Models (VLMs) have garnered increasing attention in the AI community due to their promising practical applications.<n>Recent studies attribute these hallucinations to VLMs' over-reliance on linguistic priors and insufficient visual feature integration.<n>We propose an adversarial parametric editing framework for Hallucination mitigation inVLMs, which follows an textbfActtextbfLocate-textbfEdit textbfAdversarially paradigm.
  arXiv  Detail & Related papers  (2025-12-26T11:56:45Z)
• PruneHal: Reducing Hallucinations in Multi-modal Large Language Models through Adaptive KV Cache Pruning [87.35309934860938]
  hallucinations in large language models (MLLMs) are strongly associated with insufficient attention allocated to visual tokens.<n>We propose textbfPruneHal, a training-free, simple yet effective method that leverages adaptive KV cache pruning to enhance the model's focus on critical visual information.
  arXiv  Detail & Related papers  (2025-10-22T02:41:07Z)
• MIRAGE: Assessing Hallucination in Multimodal Reasoning Chains of MLLM [58.2298313720146]
  Multimodal hallucinations are multi-sourced and arise from diverse causes.<n>Existing benchmarks fail to adequately distinguish between perception-induced hallucinations and reasoning-induced hallucinations.
  arXiv  Detail & Related papers  (2025-05-30T05:54:36Z)
• Grounding Language with Vision: A Conditional Mutual Information Calibrated Decoding Strategy for Reducing Hallucinations in LVLMs [51.93737995405164]
  Large Vision-Language Models (LVLMs) are susceptible to hallucinations.<n>We introduce a novel Conditional Pointwise Mutual Information (C-PMI) calibrated decoding strategy.<n>We show that the proposed method significantly reduces hallucinations in LVLMs while preserving decoding efficiency.
  arXiv  Detail & Related papers  (2025-05-26T08:36:10Z)
• Decoupling Contrastive Decoding: Robust Hallucination Mitigation in Multimodal Large Language Models [23.846371966667927]
  multimodal large language models (MLLMs) generate outputs misaligned with obvious visual or factual evidence.<n>We propose a novel framework: Decoupling Contrastive Decoding (DCD)<n>DCD decouples the learning of positive and negative samples in preference datasets, and trains separate positive and negative image projections within the MLLM.
  arXiv  Detail & Related papers  (2025-04-09T02:59:18Z)
• Mitigating Hallucinations in Large Vision-Language Models with Instruction Contrastive Decoding [25.489832294197797]
  This paper introduces the Instruction Contrastive Decoding (ICD) method, a novel approach designed to reduce hallucinations during LVLM inference. Our method is inspired by our observation that what we call disturbance instructions significantly exacerbate hallucinations in multimodal fusion modules.
  arXiv  Detail & Related papers  (2024-03-27T16:04:47Z)

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.