Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2504.13169 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908601066455040 |
|---|---|
| author | Wu, Tsung-Han Lee, Heekyung Ge, Jiaxin Gonzalez, Joseph E. Darrell, Trevor Chan, David M. |
| author_facet | Wu, Tsung-Han Lee, Heekyung Ge, Jiaxin Gonzalez, Joseph E. Darrell, Trevor Chan, David M. |
| contents | Vision-Language Models (VLMs) excel at visual understanding but often suffer from visual hallucinations, where they generate descriptions of nonexistent objects, actions, or concepts, posing significant risks in safety-critical applications. Existing hallucination mitigation methods typically follow one of two paradigms: generation adjustment, which modifies decoding behavior to align text with visual inputs, and post-hoc verification, where external models assess and correct outputs. While effective, generation adjustment methods often rely on heuristics and lack correction mechanisms, while post-hoc verification is complicated, typically requiring multiple models and tending to reject outputs rather than refine them. In this work, we introduce REVERSE, a unified framework that integrates hallucination-aware training with on-the-fly self-verification. By leveraging a new hallucination-verification dataset containing over 1.3M semi-synthetic samples, along with a novel inference-time retrospective resampling technique, our approach enables VLMs to both detect hallucinations during generation and dynamically revise those hallucinations. Our evaluations show that REVERSE achieves state-of-the-art hallucination reduction, outperforming the best existing methods by up to 12% on CHAIR-MSCOCO and 34% on HaloQuest. Our dataset, model, and code are available at: https://reverse-vlm.github.io. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_13169 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Generate, but Verify: Reducing Hallucination in Vision-Language Models with Retrospective Resampling Wu, Tsung-Han Lee, Heekyung Ge, Jiaxin Gonzalez, Joseph E. Darrell, Trevor Chan, David M. Computer Vision and Pattern Recognition Vision-Language Models (VLMs) excel at visual understanding but often suffer from visual hallucinations, where they generate descriptions of nonexistent objects, actions, or concepts, posing significant risks in safety-critical applications. Existing hallucination mitigation methods typically follow one of two paradigms: generation adjustment, which modifies decoding behavior to align text with visual inputs, and post-hoc verification, where external models assess and correct outputs. While effective, generation adjustment methods often rely on heuristics and lack correction mechanisms, while post-hoc verification is complicated, typically requiring multiple models and tending to reject outputs rather than refine them. In this work, we introduce REVERSE, a unified framework that integrates hallucination-aware training with on-the-fly self-verification. By leveraging a new hallucination-verification dataset containing over 1.3M semi-synthetic samples, along with a novel inference-time retrospective resampling technique, our approach enables VLMs to both detect hallucinations during generation and dynamically revise those hallucinations. Our evaluations show that REVERSE achieves state-of-the-art hallucination reduction, outperforming the best existing methods by up to 12% on CHAIR-MSCOCO and 34% on HaloQuest. Our dataset, model, and code are available at: https://reverse-vlm.github.io. |
| title | Generate, but Verify: Reducing Hallucination in Vision-Language Models with Retrospective Resampling |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2504.13169 |