arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2405.16282cs.CLcs.AIcs.LG
收藏

引擎盖下的置信度:对大语言模型中置信度-概率对齐性的研究

Confidence Under the Hood: An Investigation into the Confidence-Probability Alignment in Large Language Models

Brock University(布洛克大学);Nazarbayev University(纳扎尔巴耶夫大学);Massachusetts Institute of Technology(麻省理工学院)

Abhishek Kumar, Robert Morabito, Sanzhar Umbet, Jad Kabbara, Ali Emami

AI总结:

本文提出置信度-概率对齐性概念,通过多种提示技术探究LLM内部token概率置信度与显式表达置信度的一致性,发现GPT-4对齐性最强,为LLM风险评估与可信度研究提供支持。

摘要:

随着大语言模型(LLMs)的使用日益广泛,理解它们对生成回答的置信度自我评估变得愈发重要,因为这直接关系到这些模型输出的可靠性。我们引入了置信度-概率对齐性(Confidence-Probability Alignment)这一概念,它将LLM的内部置信度(通过token概率量化)与模型在被明确问及其确定性时在回答中所表达的置信度联系起来。通过使用多种数据集和鼓励模型内省(introspection)的提示技术,我们探究了模型内部置信度与表达置信度之间的对齐情况。这些技术包括使用结构化评估量表来评定置信度、在提示时纳入答案选项,以及引导模型对其不认为是自己生成的输出给出置信度水平。值得注意的是,在所分析的模型中,OpenAI的GPT-4表现出最强的置信度-概率对齐性,在广泛的任务中平均Spearman's $\hatρ$达到0.42。我们的工作有助于推动LLM应用中的风险评估工作,并加深我们对模型可信度的理解。

英文摘要:

As the use of Large Language Models (LLMs) becomes more widespread, understanding their self-evaluation of confidence in generated responses becomes increasingly important as it is integral to the reliability of the output of these models. We introduce the concept of Confidence-Probability Alignment, that connects an LLM's internal confidence, quantified by token probabilities, to the confidence conveyed in the model's response when explicitly asked about its certainty. Using various datasets and prompting techniques that encourage model introspection, we probe the alignment between models' internal and expressed confidence. These techniques encompass using structured evaluation scales to rate confidence, including answer options when prompting, and eliciting the model's confidence level for outputs it does not recognize as its own. Notably, among the models analyzed, OpenAI's GPT-4 showed the strongest confidence-probability alignment, with an average Spearman's $\hatρ$ of 0.42, across a wide range of tasks. Our work contributes to the ongoing efforts to facilitate risk assessment in the application of LLMs and to further our understanding of model trustworthiness.

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。

↑