Cracking the Code of Hallucination in LVLMs with Vision-aware Head Divergence
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916786301042688 |
|---|---|
| author | He, Jinghan Zhu, Kuan Guo, Haiyun Fang, Junfeng Hua, Zhenglin Jia, Yuheng Tang, Ming Chua, Tat-Seng Wang, Jinqiao |
| author_facet | He, Jinghan Zhu, Kuan Guo, Haiyun Fang, Junfeng Hua, Zhenglin Jia, Yuheng Tang, Ming Chua, Tat-Seng Wang, Jinqiao |
| contents | Large vision-language models (LVLMs) have made substantial progress in integrating large language models (LLMs) with visual inputs, enabling advanced multimodal reasoning. Despite their success, a persistent challenge is hallucination-where generated text fails to accurately reflect visual content-undermining both accuracy and reliability. Existing methods focus on alignment training or decoding refinements but primarily address symptoms at the generation stage without probing the underlying causes. In this work, we investigate the internal mechanisms driving hallucination in LVLMs, with an emphasis on the multi-head attention module. Specifically, we introduce Vision-aware Head Divergence (VHD), a metric that quantifies the sensitivity of attention head outputs to visual context. Based on this, our findings reveal the presence of vision-aware attention heads that are more attuned to visual information; however, the model's overreliance on its prior language patterns is closely related to hallucinations. Building on these insights, we propose Vision-aware Head Reinforcement (VHR), a training-free approach to mitigate hallucination by enhancing the role of vision-aware attention heads. Extensive experiments demonstrate that our method achieves superior performance compared to state-of-the-art approaches in mitigating hallucinations, while maintaining high efficiency with negligible additional time overhead. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_13949 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Cracking the Code of Hallucination in LVLMs with Vision-aware Head Divergence He, Jinghan Zhu, Kuan Guo, Haiyun Fang, Junfeng Hua, Zhenglin Jia, Yuheng Tang, Ming Chua, Tat-Seng Wang, Jinqiao Computation and Language Computer Vision and Pattern Recognition Large vision-language models (LVLMs) have made substantial progress in integrating large language models (LLMs) with visual inputs, enabling advanced multimodal reasoning. Despite their success, a persistent challenge is hallucination-where generated text fails to accurately reflect visual content-undermining both accuracy and reliability. Existing methods focus on alignment training or decoding refinements but primarily address symptoms at the generation stage without probing the underlying causes. In this work, we investigate the internal mechanisms driving hallucination in LVLMs, with an emphasis on the multi-head attention module. Specifically, we introduce Vision-aware Head Divergence (VHD), a metric that quantifies the sensitivity of attention head outputs to visual context. Based on this, our findings reveal the presence of vision-aware attention heads that are more attuned to visual information; however, the model's overreliance on its prior language patterns is closely related to hallucinations. Building on these insights, we propose Vision-aware Head Reinforcement (VHR), a training-free approach to mitigate hallucination by enhancing the role of vision-aware attention heads. Extensive experiments demonstrate that our method achieves superior performance compared to state-of-the-art approaches in mitigating hallucinations, while maintaining high efficiency with negligible additional time overhead. |
| title | Cracking the Code of Hallucination in LVLMs with Vision-aware Head Divergence |
| topic | Computation and Language Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2412.13949 |