SRLoRA replaces low-importance LoRA pairs with unused SVD directions during fine-tuning, keeping the trainable parameter count constant and aiming for faster, better adaptation.
DeBERTav3: Improving deBERTa using ELECTRA-style pre-training with gradient-disentangled embedding sharing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
REJECT 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
SRLoRA: Subspace Recomposition in Low-Rank Adaptation via Importance-Based Fusion and Reinitialization
SRLoRA replaces low-importance LoRA pairs with unused SVD directions during fine-tuning, keeping the trainable parameter count constant and aiming for faster, better adaptation.