BriefGPT.xyz
Aug, 2024
公平多模态大语言模型的社会去偏见
Social Debiasing for Fair Multi-modal LLMs
HTML
PDF
Harry Cheng, Yangyang Guo, Qingpei Guo, Ming Yang, Tian Gan...
TL;DR
本研究解决了多模态大语言模型（MLLMs）中的社会偏见问题，提出了一种综合的反事实数据集（CMSC）以及去偏见策略（ASD）。通过改进训练过程和数据采样方法，实验结果表明，该方法显著降低了社会偏见，同时保持了模型的原始性能。
Abstract
Multi-modal
Large Language Models
(MLLMs) have advanced significantly, offering powerful vision-language understanding capabilities. However, these models often inherit severe social biases from their training da
→