Mitigating Hallucination in Multimodal Large Language Model via Hallucination-targeted Direct Preference Optimization
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910700085968896 |
|---|---|
| author | Fu, Yuhan Xie, Ruobing Sun, Xingwu Kang, Zhanhui Li, Xirong |
| author_facet | Fu, Yuhan Xie, Ruobing Sun, Xingwu Kang, Zhanhui Li, Xirong |
| contents | Multimodal Large Language Models (MLLMs) are known to hallucinate, which limits their practical applications. Recent works have attempted to apply Direct Preference Optimization (DPO) to enhance the performance of MLLMs, but have shown inconsistent improvements in mitigating hallucinations. To address this issue more effectively, we introduce Hallucination-targeted Direct Preference Optimization (HDPO) to reduce hallucinations in MLLMs. Unlike previous approaches, our method tackles hallucinations from their diverse forms and causes. Specifically, we develop three types of preference pair data targeting the following causes of MLLM hallucinations: (1) insufficient visual capabilities, (2) long context generation, and (3) multimodal conflicts. Experimental results demonstrate that our method achieves superior performance across multiple hallucination evaluation datasets, surpassing most state-of-the-art (SOTA) methods and highlighting the potential of our approach. Ablation studies and in-depth analyses further confirm the effectiveness of our method and suggest the potential for further improvements through scaling up. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2411_10436 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Mitigating Hallucination in Multimodal Large Language Model via Hallucination-targeted Direct Preference Optimization Fu, Yuhan Xie, Ruobing Sun, Xingwu Kang, Zhanhui Li, Xirong Computation and Language Artificial Intelligence Computer Vision and Pattern Recognition Multimedia Multimodal Large Language Models (MLLMs) are known to hallucinate, which limits their practical applications. Recent works have attempted to apply Direct Preference Optimization (DPO) to enhance the performance of MLLMs, but have shown inconsistent improvements in mitigating hallucinations. To address this issue more effectively, we introduce Hallucination-targeted Direct Preference Optimization (HDPO) to reduce hallucinations in MLLMs. Unlike previous approaches, our method tackles hallucinations from their diverse forms and causes. Specifically, we develop three types of preference pair data targeting the following causes of MLLM hallucinations: (1) insufficient visual capabilities, (2) long context generation, and (3) multimodal conflicts. Experimental results demonstrate that our method achieves superior performance across multiple hallucination evaluation datasets, surpassing most state-of-the-art (SOTA) methods and highlighting the potential of our approach. Ablation studies and in-depth analyses further confirm the effectiveness of our method and suggest the potential for further improvements through scaling up. |
| title | Mitigating Hallucination in Multimodal Large Language Model via Hallucination-targeted Direct Preference Optimization |
| topic | Computation and Language Artificial Intelligence Computer Vision and Pattern Recognition Multimedia |
| url | https://arxiv.org/abs/2411.10436 |