Bayesian Low-rank Adaptation for Large Language Models
2023-08-24Code Available1· sign in to hype
Adam X. Yang, Maxime Robeyns, Xi Wang, Laurence Aitchison
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/maximerobeyns/bayesian_loraOfficialIn paperpytorch★ 38
- github.com/adamxyang/laplace-loraOfficialIn paperpytorch★ 28
Abstract
Low-rank adaptation (LoRA) has emerged as a new paradigm for cost-efficient fine-tuning of large language models (LLMs). However, fine-tuned LLMs often become overconfident especially when fine-tuned on small datasets. Bayesian methods, with their inherent ability to estimate uncertainty, serve as potent tools to mitigate overconfidence and enhance calibration. In this work, we introduce Laplace-LoRA, which applies a Bayesian approach to the LoRA parameters. Specifically, Laplace-LoRA applies a Laplace approximation to the posterior over the LoRA parameters, considerably improving the calibration of fine-tuned LLMs.