Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Learning from Fine-Grained Visual Discrepancies: Mitigating Multimodal Hallucinations via In-Context Visual Contrastive Optimization

About

Multimodal hallucination remains a persistent challenge for Vision-Language Models (VLMs). Standard textual Direct Preference Optimization (DPO) often fails to mitigate it due to a lack of explicit visual supervision. While existing works introduce visual preference DPO by contrasting original images against negative ones, they suffer from a theoretically inconsistent objective caused by partition function mismatches and rely on coarse-grained negatives that could enable shortcut learning. In this work, we propose In-Context Visual Contrastive Optimization (IC-VCO). By placing contrastive images within a shared multi-image context, IC-VCO ensures a mathematically rigorous objective. We further introduce Visual Contrast Distillation (VCDist), an auxiliary reliability-gated regularizer that encourages consistency between multi-image contrastive training and single-image inference. Finally, we propose a contrastive sample editing strategy that generates hard negatives via precise semantic perturbations. Experiments on five benchmarks demonstrate IC-VCO's best overall performance and the effectiveness of our sample editing strategy. Code and data are available at https://github.com/OPPO-Mente-Lab/IC-VCO.

Haolin Deng, Xin Zou, Zhiwei Jin, Chen Chen, Haonan Lu, Xuming Hu• 2026

Related benchmarks

TaskDatasetResultRank
Visual PerceptionBLINK--
241
Hallucination EvaluationHallusionBench--
153
Visual Hallucination EvaluationHallusionBench--
120
Visual Perception and ReasoningBLINK
Accuracy49.01
64
Multi-modal Hallucination EvaluationAMBER--
28
Compositional Reasoning and Perception EvaluationCRPE
Exist Score94.15
13
General Multimodal EvaluationMacro-average of HallusionBench, AMBER, CRPE, R-Bench, and BLINK
Overall Score63.35
13
Multimodal Hallucination EvaluationCRPE
Existence Score96.8
13
Multimodal Hallucination EvaluationR-Bench
Dis66.68
13
RobustnessR-Bench
R-Bench Dis Metric60.71
13
Showing 10 of 10 rows

Other info

Follow for update