BaLoRA: Bayesian Low-Rank Adaptation of Large Scale Models

Dario Coscia, Sindy Löwe, Max Welling

arXiv:2605.08110 · 2026-05-12 공개 · arXiv · PDF

vision-language uncertainty-quantification low-rank-adaptation parameter-efficiency natural-language-reasoning model-fine-tuning zero-shot-inference balo-ra

Abstract

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.

한국어 요약

📋 한 줄 요약

**[효율적 미세조정 / 베이지안 LoRA]** 입력 적응적 베이지안 파라미터화로 LoRA에 불확실성 정량화와 정확도 향상을 동시에 부여한 BaLoRA 제안.

🎯 핵심 기여도

💡 핵심 아이디어

적응적 잡음 주입(adaptive noise injection)은 정칙화 효과로 작용해 LoRA의 표현력 부족을 보완하며, 동시에 베이지안 사후 분포로 자연스럽게 해석돼 불확실성도 제공한다는 통합 시각.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: 신뢰성이 정확도만큼 중요한 과학·의료·소재 발견 영역에서 LoRA를 실제로 쓸 수 있게 만드는 실용적 베이지안 확장. **한계**: 추론 시 샘플링으로 인한 비용 증가, 변분 가정에 따른 사후 근사 오차.

🚀 실용적 활용