The Illusion of Progress? A Critical Look at Test-Time Adaptation for Vision-Language Models
This research critically examines the reliability of test-time adaptation in vision-language models, introducing the TTA-VLM benchmark to address current inc...