GS-Bias improves zero-shot CLIP accuracy by learning per-image global and spatial biases at the output-score level, achieving reported state-of-the-art test-time adaptation with far lower memory use than TPT.
Describing textures in the wild
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
GS-Bias: Global-Spatial Bias Learner for Single-Image Test-Time Adaptation of Vision-Language Models
GS-Bias improves zero-shot CLIP accuracy by learning per-image global and spatial biases at the output-score level, achieving reported state-of-the-art test-time adaptation with far lower memory use than TPT.