Large vision language models stop seeing things that are not there
From Suppression to Repair: Mitigating Object Hallucination in Large Vision-Language Models via Localized Distribution Alignment
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Large vision-language models sometimes make mistakes by imagining objects that aren't really in the picture, called hallucinations. The authors introduce a way to fix this problem without retraining the model, by adjusting its internal representations at the time it makes predictions. They move the parts of the model's memory related to hallucinated objects closer to more accurate parts, reducing errors while keeping the model’s abilities strong. Tests show this method improves both accuracy and the quality of image descriptions.
What this means in practice
- •For software developers: Improve the accuracy of image descriptions generated by vision-language AI without retraining the entire model.
- •For multimedia content creators: Create more reliable AI-generated captions and descriptions for images that reduce misleading or false object mentions.
Authors
Chen Zhao, Xingping Dong, Jiachun Shi, Liang Peng, Chong Wang, Zhen Lei, Ran He, Bo Du
Abstract
Object hallucination remains a major obstacle for large vision-language models (LVLMs) to generate reliable content. An intuitive mitigation strategy is to suppress hallucination-related components in hidden representations. However, these components may also contain useful information, and suppressing them can weaken the model's multimodal capabilities. In this paper, we propose ResOT, a training-free method that repairs representations at inference time through localized distribution alignment. Specifically, ResOT projects dominant hallucinated directions away from the faithful subspace, forming a low-dimensional residual subspace for intervention. Within this subspace, ResOT uses Gaussian optimal transport (OT) to align the hallucinated distribution with the faithful one. The resulting map defines repair targets with minimal changes to the original representations. At inference, ResOT adaptively controls how far each token state moves toward its OT target. Experiments on three representative LVLMs show that ResOT substantially reduces object hallucination while improving image caption quality and multimodal performance across multiple benchmarks. Code will be released.