Learnable Rank Improves LoRA Fine-Tuning

Arxiv CS.CL (NLP/language models)

Paper introduces learnable rank in LoRA adapters, removing fixed low-rank bias. It achieves better performance-efficiency trade-offs on benchmarks.

lorafine-tuning

Arxiv CS.CL (NLP/language models)·Arpit GargSimon LuceyHemanth Saratchandran

7 days ago