Freezing the most important ViT parameter matrices before each new task, on top of orthogonal LoRA composition, reduces forgetting and improves average accuracy in class-incremental learning.
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
-
LoRA-Based Continual Learning with Constraints on Critical Parameter Changes
Freezing the most important ViT parameter matrices before each new task, on top of orthogonal LoRA composition, reduces forgetting and improves average accuracy in class-incremental learning.