Uncertainty-Based Abstention in LLMs Improves Safety and Reduces Hallucinations
Keywords: LLMs, uncertainty, abstention, correctness, hallucinations, safety
TL;DR: Abstention based on the right form of uncertainty improves correctness, hallucinations and safety in LLMs.
Abstract: A major barrier to the practical deployment of large language models (LLMs) is their lack of reliability. Three situations where this is particularly apparent are correctness, hallucinations when given unanswerable questions, and safety where responses are harmful or offensive. In all three cases, models should ideally abstain from responding---much like humans refrain from answering questions when uncertain. Inspired by analogous approaches in classification, this study explores the feasibility and efficacy of LLMs abstaining when uncertain in the domain of question-answering. We investigate two kinds of uncertainties, statistical uncertainty metrics and a distinct verbalized measure, termed as In Dialogue Uncertainty (InDU), measuring hedge words such as `I don't know' in responses. Using these uncertainty measures combined with models with and without reinforcement learning with human feedback (RLHF), we show in all three situations, abstention based on the right kind of uncertainty measure can boost the reliability of LLMs. By abstaining for a few highly uncertain samples we improve correctness by up to 8\%, avoid 50\% of hallucinations by correctly identifying unanswerable questions, and in particular increase safety by 70-99\% with almost no additional computational overhead.
Primary Area: alignment, fairness, safety, privacy, and societal considerations
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13491
Loading