将大模型不确定性拆解为三类，帮人精准定位错误根源。
The Anatomy of Uncertainty in LLMs
- 把不确定性的来源分为提示模糊、知识不足和采样随机三类。
- 实验发现不同模型规模和任务下，主导因素会变化。
- 适合想改进模型可靠性和检测幻觉的研究者使用。
理解大语言模型（LLM）对回答不确定的原因，对其可靠部署至关重要。现有方法或仅提供单一不确定度分数，或依赖经典的偶然性-认知性二分法，难以提供可操作的改进洞察。近期研究也表明，这些方法不足以解释LLM中的不确定性。本文提出一种不确定性分解框架，将LLM的不确定性拆分为三个语义层面：（i）输入模糊性，源于提示不明确；（ii）知识缺口，由参数证据不足引起；（iii）解码随机性，来自随机采样。通过一系列实验，我们发现这些成分的主导地位会随模型规模和任务类型而变化。该框架有助于更深入审计模型可靠性并检测幻觉，为针对性干预和构建更可信系统铺平道路。
原文摘要 · Abstract (English)
Understanding why a large language model (LLM) is uncertain about the response is important for their reliable deployment. Current approaches, which either provide a single uncertainty score or rely on the classical aleatoric-epistemic dichotomy, fail to offer actionable insights for improving the generative model. Recent studies have also shown that such methods are not enough for understanding uncertainty in LLMs. In this work, we advocate for an uncertainty decomposition framework that dissects LLM uncertainty into three distinct semantic components: (i) input ambiguity, arising from ambiguous prompts; (ii) knowledge gaps, caused by insufficient parametric evidence; and (iii) decoding randomness, stemming from stochastic sampling. Through a series of experiments we demonstrate that the dominance of these components can shift across model size and task. Our framework provides a better understanding to audit LLM reliability and detect hallucinations, paving the way for targeted interventions and more trustworthy systems.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。