3

Low-rank finetuning for LLMs: A fairness perspective