Seeing Far and Clearly: Mitigating Hallucinations in MLLMs with Attention Causal Decoding
Tang, Feilong Liu, Chengzhi Xu, Zhongxing Hu, Ming Chen, Ziyang Peng, Zelin Yang, Zhiwei Zhou, Sijin
Tang, Feilong
Liu, Chengzhi
Xu, Zhongxing
Hu, Ming
Chen, Ziyang
Peng, Zelin
Yang, Zhiwei
Zhou, Sijin
Author
Supervisor
Department
Computational Biology
Embargo End Date
Type
Conference proceeding
Date
2025
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Recent advancements in multimodal large language models (MLLMs) have significantly improved performance in visual question answering. However, they often suffer from hallucinations. In this work, hallucinations are categorized into two main types: initial hallucinations and snowball hallucinations. We argue that adequate contextual information can be extracted directly from the token interaction process. Inspired by causal inference in the decoding strategy, we propose to leverage causal masks to establish information propagation between multimodal tokens. The hypothesis is that insufficient interaction between those tokens may lead the model to rely on outlier tokens, overlooking dense and rich contextual cues. Therefore, we propose to intervene in the propagation process by tackling outlier tokens to enhance in-context inference. With this goal, we present FarSight, a versatile plug-and-play decoding strategy to reduce attention interference from outlier tokens merely by optimizing the causal mask. The heart of our method is effective token propagation. We design an attention register structure within the upper triangular matrix of the causal mask, dynamically allocating attention to capture attention diverted to outlier tokens. Moreover, a positional awareness encoding method with a diminishing masking rate is proposed, allowing the model to attend to further preceding tokens, especially for video sequence tasks.
Citation
F. Tang et al., "Seeing Far and Clearly: Mitigating Hallucinations in MLLMs with Attention Causal Decoding," 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA, 2025, pp. 26147-26159, doi: 10.1109/CVPR52734.2025.02435.
Source
Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
Conference
2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025
Keywords
Farsight Demonstrates Significant Hallucination-mitigating Performance Across Different Mllms On Both Image And Video Benchmarks, Proving Its Effectiveness, With Extensive Experiments
Subjects
Source
2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025
Publisher
IEEE