arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.02543cs.LGcs.AIcs.CL
收藏

相信还是不相信你的大语言模型

To Believe or Not to Believe Your LLM

Google DeepMind（谷歌DeepMind）

Yasin Abbasi Yadkori, Ilja Kuzborskij, András György, Csaba Szepesvári

阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出一种基于信息论的不确定性量化指标，通过迭代提示检测大语言模型在单答案和多答案场景中的幻觉，优于标准策略。

摘要：

我们探索大语言模型（LLMs）中的不确定性量化，目标是识别何时给定查询的响应不确定性较大。我们同时考虑认知不确定性和偶然不确定性，前者源于对真实情况（如事实或语言）缺乏知识，后者源于不可约的随机性（如多种可能的答案）。特别是，我们推导出一个信息论指标，能够可靠地检测何时仅认知不确定性较大，此时模型输出不可靠。该条件可以仅基于模型输出计算，而模型输出仅需通过基于先前响应的特殊迭代提示获得。这种量化，例如，允许在单答案和多答案响应中检测幻觉（认知不确定性较高的情况）。这与许多标准不确定性量化策略（如对响应的对数似然设置阈值）形成对比，在多答案情况下无法检测幻觉。我们进行了一系列实验，证明了我们公式的优势。此外，我们的研究揭示了迭代提示如何放大LLM分配给给定输出的概率，这可能具有独立的研究价值。

英文摘要：

We explore uncertainty quantification in large language models (LLMs), with the goal to identify when uncertainty in responses given a query is large. We simultaneously consider both epistemic and aleatoric uncertainties, where the former comes from the lack of knowledge about the ground truth (such as about facts or the language), and the latter comes from irreducible randomness (such as multiple possible answers). In particular, we derive an information-theoretic metric that allows to reliably detect when only epistemic uncertainty is large, in which case the output of the model is unreliable. This condition can be computed based solely on the output of the model obtained simply by some special iterative prompting based on the previous responses. Such quantification, for instance, allows to detect hallucinations (cases when epistemic uncertainty is high) in both single- and multi-answer responses. This is in contrast to many standard uncertainty quantification strategies (such as thresholding the log-likelihood of a response) where hallucinations in the multi-answer case cannot be detected. We conduct a series of experiments which demonstrate the advantage of our formulation. Further, our investigations shed some light on how the probabilities assigned to a given output by an LLM can be amplified by iterative prompting, which might be of independent interest.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

↑