Reducing Hallucination in Vision-Language Models via Stage-wise Preference Optimization under Distribution Shift

• URL: http://arxiv.org/abs/2605.16411v1
• Date: Wed, 13 May 2026 15:37:51 GMT
• Title: Reducing Hallucination in Vision-Language Models via Stage-wise Preference Optimization under Distribution Shift
• Abstract summary: Hallucination remains a fundamental challenge in vision-language models (VLMs)<n>We propose a stage-wise preference optimization framework for hallucination reduction through targeted multimodal constructions.<n>Experiments on open-source benchmarks and real-world multimodal evaluation scenarios demonstrate improved grounding consistency, reduced hallucination, and more informative grounded responses.
• Score: 0.0
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Hallucination remains a fundamental challenge in vision-language models (VLMs), where autoregressive generation may produce linguistically plausible yet physically inconsistent or visually ungrounded responses due to likelihood maximization under joint probabilistic modeling. We propose a stage-wise preference optimization framework for hallucination reduction through targeted multimodal data construction. Rather than directly optimizing on generic instruction-following data, our approach progressively constructs hallucination-focused preference pairs near known failure boundaries. The framework emphasizes ambiguous spatial orientation, object relationships, OCR uncertainty, and adversarial false-premise training. Hallucinated negatives are generated through minimally perturbed yet visually inconsistent alternatives, enabling Direct Preference Optimization (DPO) to better separate grounded reasoning from plausible hallucination. Experiments on open-source benchmarks and real-world multimodal evaluation scenarios demonstrate improved grounding consistency, reduced hallucination, and more informative grounded responses. Cross-model qualitative evaluation further shows that the proposed multimodal LLM DPO framework produces more visually grounded responses than several frontier proprietary VLMs, such as in ambiguous spatial reasoning and adversarial false-premise settings. The results suggest that hallucination may arise not only from limited model capacity, but also from inherent tendencies of autoregressive probabilistic generation to favor linguistically plausible continuations under weak visual grounding. Future work may explore physical consistency modeling, uncertainty-aware multimodal reasoning, and architectural alternatives beyond standard autoregressive decoding.

Related papers

• Beyond Global Editing: Per-Instance Disentangled Subspaces for Training-Free Hallucination Mitigation in LVLMs [16.83614016603383]
  We propose a training-free hallucination mitigation framework for large vision-language models.<n>Our method first constructs a set of Disentangled Hallucination Subspaces, each isolating a distinct hallucination mode.<n>Experiments across multiple vision-language benchmarks and LVLM families demonstrate consistent improvements.
  arXiv  Detail & Related papers  (2026-08-10T09:21:12Z)
• Clearer Sight, Fewer Lies: Oriented Pickup Preference Optimization for Multimodal Hallucination Mitigation [41.3823869123186]
  Multimodal Large Language Models (MLLMs) are prone to hallucination as their generation preferences are insufficiently calibrated to visual evidence.<n>We propose Oriented Pickup Preference Optimization (textttOPPO), an evidence-aware alignment objective that learns preferences over the strength of visual evidence.
  arXiv  Detail & Related papers  (2026-06-29T05:33:22Z)
• 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)
• 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)
• Online Self-Calibration Against Hallucination in Vision-Language Models [23.13137973421435]
  Large Vision-Language Models (LVLMs) often suffer from hallucinations, generating descriptions that include visual details absent from the input image.<n>We propose textbfOnline textbfSelf-textbfCAlibtextbfRation (OSCAR), a framework that integrates Monte Carlo Tree Search with a Dual-Granularity Reward Mechanism to construct preference data.
  arXiv  Detail & Related papers  (2026-05-01T01:03:05Z)
• Decoding by Perturbation: Mitigating MLLM Hallucinations via Dynamic Textual Perturbation [36.57014987764294]
  Decoding by Perturbation (DeP) is a training-free framework mitigating prior-induced hallucinations.<n>It employs a dynamic probe applying multi-level textual perturbations to elicit latent language priors.<n>Extensive experiments confirm DeP effectively reduces hallucinations and achieves superior performance across multiple benchmarks.
  arXiv  Detail & Related papers  (2026-04-14T08:15:44Z)
• Mitigating Entangled Steering in Large Vision-Language Models for Hallucination Reduction [49.96701537295129]
  Large Vision-Language Models (LVLMs) have achieved remarkable success across cross-modal tasks but remain hindered by hallucinations.<n>Existing methods mitigate hallucinations but often alter generation behavior, resulting in shorter outputs and shifted token distributions.<n>We propose MESA, an effective plug-and-play framework that performs controlled and selective latent intervention for hallucination mitigation.
  arXiv  Detail & Related papers  (2026-04-09T07:31:27Z)
• ACT Now: Preempting LVLM Hallucinations via Adaptive Context Integration [71.21097024566285]
  Large-Language Models (LVLMs) frequently suffer from severe hallucination issues.<n>Existing mitigation strategies rely on linguistic, single-step states to enhance visual focus or suppress strong priors.<n>We propose Adaptive Context in VisionTegration (ACT), a training-free inference intervention method that mitigates hallucination through the adaptive integration of contextual information.
  arXiv  Detail & Related papers  (2026-04-01T14:49:50Z)
• MoD-DPO: Towards Mitigating Cross-modal Hallucinations in Omni LLMs using Modality Decoupled Preference Optimization [4.088161686930475]
  We propose Modality-Decoupled Direct Preference Optimization (MoD-DPO) for improving modality grounding in omni LLMs.<n>MoD-DPO introduces modality-aware regularization terms that explicitly enforce invariance to corruptions in irrelevant modalities and sensitivity to perturbations in relevant modalities.<n>Experiments demonstrate that MoD-DPO consistently improves perception accuracy and hallucination resistance, outperforming previous preference optimization baselines.
  arXiv  Detail & Related papers  (2026-03-03T17:50:24Z)
• Seeing Through the Chain: Mitigate Hallucination in Multimodal Reasoning Models via CoT Compression and Contrastive Preference Optimization [78.94590726578014]
  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.
  arXiv  Detail & Related papers  (2026-02-03T11:00:55Z)
• Mitigating Hallucinations in Large Vision-Language Models by Self-Injecting Hallucinations [73.37711261605271]
  hallucination mitigation methods are mainly based on preference alignment and require external human annotations or auxiliary models for preference data collection.<n>We propose Autonomous Preference Alignment via Self-Injection (APASI), a novel and generalizable method that mitigates hallucinations without external dependencies.<n>APASI leverages the target LVLM to self-inject hallucinations into a generated response, creating a pair of responses with varying preference levels.
  arXiv  Detail & Related papers  (2025-09-14T14:26:53Z)
• Mitigating Hallucinations in Large Vision-Language Models via Entity-Centric Multimodal Preference Optimization [65.12217781259525]
  Existing preference alignment methods focus on aligning model responses with human preferences.<n>We propose Entity-centric Multimodal Preference Optimization (EMPO), which achieves enhanced modality alignment.<n> EMPO reduces hallucinations rates by 85.9% on Object-HalBench and 49.8% on MM-HalBench.
  arXiv  Detail & Related papers  (2025-06-04T15:03:50Z)
• OViP: Online Vision-Language Preference Learning for VLM Hallucination [44.14029765850719]
  Large vision-language models (LVLMs) remain vulnerable to hallucination, often generating content misaligned with visual inputs.<n>We propose an Online Vision-language Preference Learning framework that dynamically constructs contrastive training data based on the model's own hallucinations.
  arXiv  Detail & Related papers  (2025-05-21T19:26:09Z)

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.