On Fairness of Low-Rank Adaptation of Large Models

Zhoujie Ding, Ken Liu, Pura Peetathawatchai, Berivan Isik, Sanmi Koyejo

Research Area: Alignment, Evaluation, Societal implications, Compute efficient LMs, LMs for everyone

Keywords: Low-rank adaptation, LoRA, bias, fairness, subgroup fairness, evaluations, LLMs, large models

TL;DR: Compared to full-model fine-tuning, does low-rank adaptation (LoRA) have any side effects on model fairness in terms of accuracy, calibration, privacy, and gender bias? How do we know?

Abstract: Low-rank adaptation of large models, particularly LoRA, has gained traction due to its computational efficiency. This efficiency, contrasted with the prohibitive costs of full-model fine-tuning, means that practitioners often turn to LoRA without a complete understanding of its ramifications. In this study, we focus on fairness and ask whether LoRA has an unexamined impact on utility, calibration, and resistance to membership inference across different subgroups (e.g., genders, races, religions) compared to a full-model fine-tuning baseline. We present extensive experiments across vision and language domains and across classification and generation tasks using ViT-Base, Swin-v2-Large, Llama-2 7B, and Mistral 7B. Intriguingly, experiments suggest that while one can isolate cases where LoRA exacerbates model bias across subgroups, the pattern is inconsistent---in many cases, LoRA has equivalent or even improved fairness compared to the base model or its full fine-tuning baseline. We also examine the complications of evaluating fine-tuning fairness relating to task design and model token bias, calling for more careful fairness evaluations in future work.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the COLM Code of Ethics on https://colmweb.org/CoE.html

Author Guide: I certify that this submission complies with the submission instructions as described on https://colmweb.org/AuthorGuide.html

Submission Number: 32