LoRA-X is a low-rank adapter constrained to a base model's singular subspace, enabling data-free transfer to closely related models via closed-form projection.
For this comparison, we used the Oriami dataset and fine-tuned both LoRA and LoRA-X on SD-v1.5 as the source model, with SD Eff-v1.0 as the target
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-X: Bridging Foundation Models with Training-Free Cross-Model Adaptation
LoRA-X is a low-rank adapter constrained to a base model's singular subspace, enabling data-free transfer to closely related models via closed-form projection.