BaLoRA: Bayesian Low-Rank Adaptation of Large Scale Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Coscia, Dario, Löwe, Sindy, Welling, Max
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917474094546944
author Coscia, Dario
Löwe, Sindy
Welling, Max
author_facet Coscia, Dario
Löwe, Sindy
Welling, Max
contents Low-Rank Adaptation (LoRA) has become the standard for fine-tuning large pre-trained models at reduced computational cost. However, its low-rank point-estimate updates limit expressiveness, leave a persistent gap relative to full fine-tuning accuracy, and provide no built-in uncertainty quantification, limiting its applicability in settings where reliability matters as much as accuracy. We introduce BaLoRA, a Bayesian extension of LoRA with a novel input-adaptive Bayesian parameterization of LoRA matrices that adds minimal parameters and compute. Surprisingly, not only does the Bayesian extension yield well-calibrated uncertainty estimates, but the adaptive noise injection underlying our approach also significantly improves prediction accuracy, narrowing the gap with full fine-tuning across both natural language reasoning and vision tasks. When applied to band gap prediction in metal-organic frameworks, BaLoRA produces zero-shot test-time uncertainty estimates that correlate more strongly with model error than a trained ensemble of LoRA models, and improve monotonically with compute without sacrificing accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2605_08110
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle BaLoRA: Bayesian Low-Rank Adaptation of Large Scale Models
Coscia, Dario
Löwe, Sindy
Welling, Max
Machine Learning
Artificial Intelligence
Low-Rank Adaptation (LoRA) has become the standard for fine-tuning large pre-trained models at reduced computational cost. However, its low-rank point-estimate updates limit expressiveness, leave a persistent gap relative to full fine-tuning accuracy, and provide no built-in uncertainty quantification, limiting its applicability in settings where reliability matters as much as accuracy. We introduce BaLoRA, a Bayesian extension of LoRA with a novel input-adaptive Bayesian parameterization of LoRA matrices that adds minimal parameters and compute. Surprisingly, not only does the Bayesian extension yield well-calibrated uncertainty estimates, but the adaptive noise injection underlying our approach also significantly improves prediction accuracy, narrowing the gap with full fine-tuning across both natural language reasoning and vision tasks. When applied to band gap prediction in metal-organic frameworks, BaLoRA produces zero-shot test-time uncertainty estimates that correlate more strongly with model error than a trained ensemble of LoRA models, and improve monotonically with compute without sacrificing accuracy.
title BaLoRA: Bayesian Low-Rank Adaptation of Large Scale Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2605.08110