Although Large Vision-Language Models (LVLMs) have demonstrated powerful capabilities in interpreting visual information, they frequently produce content that deviates from visual information, leading to object hallucination. To tackle this, recent works mostly depend on expensive manual annotations and training cost, or significantly increase inference time. In this work, we observe that LVLMs' attention to visual information is significantly stronger when answering caption queries compared to non-caption queries. Inspired by this phenomenon, we propose Caption-sensitive Attention Intervention (CAI), a training-free, plug-and-play hallucination mitigation method that leverages the attention activation pattern in response to caption queries to enhance LVLMs' visual perception capability. Extensive experimental results across four benchmarks covering both discriminative and generative tasks, demonstrate that CAI achieves state-of-the-art (SOTA) hallucination mitigating performance only with minimal additional inference cost.
@article{arxiv.2506.23590,
title = {CAI: Caption-Sensitive Attention Intervention for Mitigating Object Hallucination in Large Vision-Language Models},
author = {Qiming Li and Zekai Ye and Xiaocheng Feng and Weihong Zhong and Libo Qin and Ruihan Chen and Baohang Li and Kui Jiang and Yaowei Wang and Ting Liu and Bing Qin},
journal= {arXiv preprint arXiv:2506.23590},
year = {2025}
}