BriefGPT.xyz
May, 2024
视觉语言模型易于执行时适应的令人沮丧的测试
Frustratingly Easy Test-Time Adaptation of Vision-Language Models
HTML
PDF
Matteo Farina, Gianni Franchi, Giovanni Iacca, Massimiliano Mancini, Elisa Ricci
TL;DR
研究表明,零温度的TTA方法(ZERO)能够在只进行一次前向传播的情况下,准确性大大超过或与现有技术相当,且速度约为10倍快,内存占用约为13倍少。
Abstract
vision-language models
seamlessly discriminate among arbitrary semantic categories, yet they still suffer from poor generalization when presented with challenging examples. For this reason,
episodic test-time adaptation
→