BriefGPT.xyz
Dec, 2022
定位潜在更新用于微调视觉语言模型
Localized Latent Updates for Fine-Tuning Vision-Language Models
HTML
PDF
Moritz Ibing, Isaak Lim, Leif Kobbelt
TL;DR
提出一种轻量级的适配器方法，通过更新预测接近观察数据点的模型以加快更新速度，并保持经典微调方法外数据的正确性，以实现在小样本学习领域中，在训练过程中看得到和看不到的类别上结果均与或优于现有技术水平。
Abstract
Although massive
pre-trained vision-language models
like
clip
show impressive generalization capabilities for many tasks, still it often remains necessary to fine-tune them for improved performance on specific da
→