arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.00367cs.CL
收藏

不要幻觉,弃权:通过多LLM协作识别大语言模型知识缺口

Don't Hallucinate, Abstain: Identifying LLM Knowledge Gaps via Multi-LLM Collaboration

University of Washington(华盛顿大学);University of California, Berkeley(加州大学伯克利分校);The Hong Kong University of Science and Technology(香港科技大学);Carnegie Mellon University(卡内基梅隆大学)

Shangbin Feng, Weijia Shi, Yike Wang, Wenxuan Ding, Vidhisha Balachandran, Yulia Tsvetkov

AI总结:

本文针对大语言模型知识缺口问题,提出基于多LLM合作与竞争协作的弃权机制,在多个问答任务上将弃权准确率提升最高19.3%,并能辅助定位检索增强失败和多跳推理知识缺口。

摘要:

尽管人们努力扩展大语言模型(LLMs)的知识,但由于知识的不断演变,知识缺口——即LLMs中缺失或过时的信息——可能始终存在。在这项工作中,我们研究了识别LLM知识缺口并在存在知识缺口时弃权(不执行)回答问题的方法。我们首先调整现有方法,通过微调/提示进行模型校准或适应,并分析它们在生成低置信度输出时弃权(不执行)的能力。鉴于这些方法在自我反思方面的失败以及对留出集的过度依赖,我们提出了两种基于模型协作的新方法,即LLMs以合作或竞争的方式相互探查知识缺口。在四个涵盖不同知识领域的问答任务上,使用三个LLMs进行的大量实验表明,揭示LLM知识缺口的合作方法和竞争方法在弃权准确率上相比最强基线实现了高达19.3%的提升。进一步分析表明,我们提出的机制有助于识别检索增强中的失败案例,并准确定位多跳推理中的知识缺口。

英文摘要:

Despite efforts to expand the knowledge of large language models (LLMs), knowledge gaps -- missing or outdated information in LLMs -- might always persist given the evolving nature of knowledge. In this work, we study approaches to identify LLM knowledge gaps and abstain from answering questions when knowledge gaps are present. We first adapt existing approaches to model calibration or adaptation through fine-tuning/prompting and analyze their ability to abstain from generating low-confidence outputs. Motivated by their failures in self-reflection and over-reliance on held-out sets, we propose two novel approaches that are based on model collaboration, i.e., LLMs probing other LLMs for knowledge gaps, either cooperatively or competitively. Extensive experiments with three LLMs on four QA tasks featuring diverse knowledge domains demonstrate that both cooperative and competitive approaches to unveiling LLM knowledge gaps achieve up to 19.3% improvements on abstain accuracy against the strongest baseline. Further analysis reveals that our proposed mechanisms could help identify failure cases in retrieval augmentation and pinpoint knowledge gaps in multi-hop reasoning.

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。

↑