EDoRA combines DoRA's magnitude-direction decomposition with LoRA-XS's frozen SVD-based sandwich parameterization, cutting trainable parameters by roughly 30x versus LoRA and DoRA on GLUE with RoBERTa-base.
A survey on multimodal large language models for autonomous driving
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
EDoRA: Efficient Weight-Decomposed Low-Rank Adaptation via Singular Value Decomposition
EDoRA combines DoRA's magnitude-direction decomposition with LoRA-XS's frozen SVD-based sandwich parameterization, cutting trainable parameters by roughly 30x versus LoRA and DoRA on GLUE with RoBERTa-base.