A CLIP adapter variant that varies per-token adaptation capacity by token importance improves new-class generalization by about one point over prior adapters.
Title resolution pending
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
-
Dynamic Rank Adaptation for Vision-Language Models
A CLIP adapter variant that varies per-token adaptation capacity by token importance improves new-class generalization by about one point over prior adapters.