Search papers, labs, and topics across Lattice.
This paper introduces Data-Adaptive Lower-Rank Adaptation (DALorRA), a variational Bayesian framework designed to enhance uncertainty estimation in large language models (LLMs) by focusing on low-rank adaptation. By implementing stochastic masking on rank dimensions, DALorRA effectively regularizes model capacity during training and improves calibration during inference, addressing the overconfidence issue prevalent in task-specific fine-tuning. Experimental results reveal that DALorRA achieves superior calibration of LLMs while maintaining high reasoning accuracy, making it a promising approach for trustworthy AI deployment.
DALorRA achieves remarkable uncertainty calibration in LLMs without sacrificing reasoning performance, tackling the critical issue of overconfidence in AI systems.
Large language models (LLMs) exhibit remarkable reasoning capabilities, but their task-specific fine-tuning is notoriously plagued by overconfidence, severely hindering trustworthy deployment. We propose Data-Adaptive Lower-Rank Adaptation (DALorRA), a simple and effective variational Bayesian sparse framework that shifts the paradigm of uncertainty quantification from the dense parameter space to the lightweight rank level of low-rank adaptation (LoRA). With the insight that LoRA essentially aggregates multiple rank-one components that may provide superfluous model capacity, DALorRA imposes stochastic masking on rank dimensions, enabling Bayesian regularization of model capacity during training and ensemble-like calibration during inference. Extensive experiments demonstrate DALorRA's excellent calibration of LLMs without compromising reasoning accuracy.
