Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2602.01047 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915885548044288 |
|---|---|
| author | Chen, Xinrong Chu, Xu Qiu, Yingmin Zhang, Hengyuan Xiong, Jing Tang, Shiyu Liu, Shuai Yang, Shaokang Yang, Cheng So, Hayden Kwok-Hay Wong, Ngai |
| author_facet | Chen, Xinrong Chu, Xu Qiu, Yingmin Zhang, Hengyuan Xiong, Jing Tang, Shiyu Liu, Shuai Yang, Shaokang Yang, Cheng So, Hayden Kwok-Hay Wong, Ngai |
| contents | Large Vision-Language Models (LVLMs) can reason from image-text inputs and perform well in various multimodal tasks. Despite this success, they are affected by language priors and often produce hallucinations. Hallucinations denote generated content that is grammatically and syntactically coherent, yet bears no match or direct relevance to visual input. To address this problem, we propose Residual Decoding (ResDec). It is a novel training-free method that uses historical information to aid decoding. The method relies on the internal implicit reasoning mechanism and token logits evolution mechanism of LVLMs to correct biases. Extensive experiments demonstrate that ResDec effectively suppresses hallucinations induced by language priors, significantly improves visual grounding, and reduces object hallucinations. In addition to mitigating hallucinations, ResDec also performs exceptionally well on comprehensive LVLM benchmarks, highlighting its broad applicability. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01047 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Residual Decoding: Mitigating Hallucinations in Large Vision-Language Models via History-Aware Residual Guidance Chen, Xinrong Chu, Xu Qiu, Yingmin Zhang, Hengyuan Xiong, Jing Tang, Shiyu Liu, Shuai Yang, Shaokang Yang, Cheng So, Hayden Kwok-Hay Wong, Ngai Computer Vision and Pattern Recognition Artificial Intelligence Large Vision-Language Models (LVLMs) can reason from image-text inputs and perform well in various multimodal tasks. Despite this success, they are affected by language priors and often produce hallucinations. Hallucinations denote generated content that is grammatically and syntactically coherent, yet bears no match or direct relevance to visual input. To address this problem, we propose Residual Decoding (ResDec). It is a novel training-free method that uses historical information to aid decoding. The method relies on the internal implicit reasoning mechanism and token logits evolution mechanism of LVLMs to correct biases. Extensive experiments demonstrate that ResDec effectively suppresses hallucinations induced by language priors, significantly improves visual grounding, and reduces object hallucinations. In addition to mitigating hallucinations, ResDec also performs exceptionally well on comprehensive LVLM benchmarks, highlighting its broad applicability. |
| title | Residual Decoding: Mitigating Hallucinations in Large Vision-Language Models via History-Aware Residual Guidance |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2602.01047 |