EUVIMEDEUVIMED entdecken
Crossref

Attribution Analysis Meets Model Editing: Advancing Knowledge Correction in Vision Language Models with VisEdit

Qizhou Chen, Taolin Zhang, Chengyu Wang, Xiaofeng He, Dakan Wang, Tingting Liu

Proceedings of the AAAI Conference on Artificial Intelligence · 11.04.2025 · 39 (2) · 2168-2176

DOI: 10.1609/aaai.v39i2.32215 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 11.04.2025. Verfasst von Qizhou Chen und 5 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Model editing aims to correct outdated or erroneous knowledge in large models without costly retraining. Recent research discovered that the mid-layer representation of the subject's final token in a prompt has a strong influence on factual predictions, and developed Large Language Model (LLM) editing techniques based on this observation. However, for Vision-LLMs (VLLMs), how visual representations impact the predictions from a decoder-only language model remains largely unexplored. To the best of our knowledge, model editing for VLLMs has not been extensively studied in the literature. In this work, we employ the contribution allocation and noise perturbation methods to measure the contributions of visual representations for token predictions. Our attribution analysis shows that visual representations in mid-to-later layers that are highly relevant to the prompt contribute significantly to predictions. Based on these insights, we propose *VisEdit*, a novel model editor for VLLMs that effectively corrects knowledge by editing intermediate visual representations in regions important to the edit prompt. We evaluated *VisEdit* using multiple VLLM backbones and public VLLM editing benchmark datasets. The results show the superiority of *VisEdit* over the strong baselines adapted from existing state-of-the-art editors for LLMs.

Zitieren

APA

Chen, Q., Zhang, T., Wang, C., He, X., Wang, D., & Liu, T. (2025). Attribution Analysis Meets Model Editing: Advancing Knowledge Correction in Vision Language Models with VisEdit. Proceedings of the AAAI Conference on Artificial Intelligence, 39(2), 2168-2176. https://doi.org/10.1609/aaai.v39i2.32215

Vancouver

Chen Q, Zhang T, Wang C, He X, Wang D, Liu T. Attribution Analysis Meets Model Editing: Advancing Knowledge Correction in Vision Language Models with VisEdit. Proceedings of the AAAI Conference on Artificial Intelligence. 2025;39(2):2168-2176. doi:10.1609/aaai.v39i2.32215

BibTeX
@article{Chen2025,
  author = {Chen, Qizhou and Zhang, Taolin and Wang, Chengyu and He, Xiaofeng and Wang, Dakan and Liu, Tingting},
  title = {Attribution Analysis Meets Model Editing: Advancing Knowledge Correction in Vision Language Models with VisEdit},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2025},
  volume = {39},
  number = {2},
  pages = {2168--2176},
  doi = {10.1609/aaai.v39i2.32215}
}
Kostenlos starten →