Who Brings the Frisbee: Probing Hidden Hallucination Factors in Large Vision-Language Model via Causality Analysis
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866908287949078528 |
|---|---|
| author | Huang, Po-Hsuan Li, Jeng-Lin Chen, Chin-Po Chang, Ming-Ching Chen, Wei-Chao |
| author_facet | Huang, Po-Hsuan Li, Jeng-Lin Chen, Chin-Po Chang, Ming-Ching Chen, Wei-Chao |
| contents | Recent advancements in large vision-language models (LVLM) have significantly enhanced their ability to comprehend visual inputs alongside natural language. However, a major challenge in their real-world application is hallucination, where LVLMs generate non-existent visual elements, eroding user trust. The underlying mechanism driving this multimodal hallucination is poorly understood. Minimal research has illuminated whether contexts such as sky, tree, or grass field involve the LVLM in hallucinating a frisbee. We hypothesize that hidden factors, such as objects, contexts, and semantic foreground-background structures, induce hallucination. This study proposes a novel causal approach: a hallucination probing system to identify these hidden factors. By analyzing the causality between images, text prompts, and network saliency, we systematically explore interventions to block these factors. Our experimental findings show that a straightforward technique based on our analysis can significantly reduce hallucinations. Additionally, our analyses indicate the potential to edit network internals to minimize hallucinated outputs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_02946 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Who Brings the Frisbee: Probing Hidden Hallucination Factors in Large Vision-Language Model via Causality Analysis Huang, Po-Hsuan Li, Jeng-Lin Chen, Chin-Po Chang, Ming-Ching Chen, Wei-Chao Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Multimedia Recent advancements in large vision-language models (LVLM) have significantly enhanced their ability to comprehend visual inputs alongside natural language. However, a major challenge in their real-world application is hallucination, where LVLMs generate non-existent visual elements, eroding user trust. The underlying mechanism driving this multimodal hallucination is poorly understood. Minimal research has illuminated whether contexts such as sky, tree, or grass field involve the LVLM in hallucinating a frisbee. We hypothesize that hidden factors, such as objects, contexts, and semantic foreground-background structures, induce hallucination. This study proposes a novel causal approach: a hallucination probing system to identify these hidden factors. By analyzing the causality between images, text prompts, and network saliency, we systematically explore interventions to block these factors. Our experimental findings show that a straightforward technique based on our analysis can significantly reduce hallucinations. Additionally, our analyses indicate the potential to edit network internals to minimize hallucinated outputs. |
| title | Who Brings the Frisbee: Probing Hidden Hallucination Factors in Large Vision-Language Model via Causality Analysis |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Multimedia |
| url | https://arxiv.org/abs/2412.02946 |