DIFO++ adapts source-free models by customizing CLIP-like vision-language models through prompt-based mutual information maximization and distilling to the target model with gap region reduction and pseudo-label fusion.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Source-Free Domain Adaptation with Vision-Language Prior
DIFO++ adapts source-free models by customizing CLIP-like vision-language models through prompt-based mutual information maximization and distilling to the target model with gap region reduction and pseudo-label fusion.