arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2408.08978cs.CL
收藏

看清LLM无法回答的问题:一种揭示LLM弱点的自我挑战框架

See What LLMs Cannot Answer: A Self-Challenge Framework for Uncovering LLM Weaknesses

Westlake University(西湖大学);University of Cambridge(剑桥大学);Microsoft GenAI(微软生成式人工智能部门);UIUC(伊利诺伊大学厄巴纳-香槟分校);Westlake Institute for Advanced Study(西湖高等研究院)

Yulong Chen, Yang Liu, Jianhao Yan, Xuefeng Bai, Ming Zhong, Yinghao Yang, Ziyi Yang, Chenguang Zhu, Yue Zhang

AI总结:

本文提出Self-Challenge框架,让GPT-4从自身错误中迭代总结错误模式并生成挑战性数据,构建SC-G4基准,揭示LLM固有弱点。

摘要:

大语言模型(LLMs)的出色表现持续超越众多人类设计的基准,这给评估LLM的不足带来了新的挑战。设计任务并发现LLM的局限性正变得越来越重要。本文研究了一个问题:LLM能否从自身所犯的错误中发现自身的局限性。为此,我们提出了一个带有人类参与的Self-Challenge评估框架。从GPT-4无法回答的种子实例出发,我们提示GPT-4总结可用于生成新实例的错误模式,并纳入人类对这些模式的反馈,以迭代地精炼这些模式,从而生成更具挑战性的数据。最终我们得到了8种多样化的模式,例如文本操作和带假设的问题。随后我们构建了一个基准SC-G4,由GPT-4利用这些模式生成的1,835个实例组成,并带有人工标注的标准答案。SC-G4作为一个具有挑战性的基准,可以对LLM的能力进行细致评估。我们的结果表明,SC-G4中只有44.96%的实例能被GPT-4正确回答。有趣的是,我们的初步研究表明,这些错误模式同样对Claude-3和Llama-3等其他LLM构成挑战,并且无法通过微调完全解决。我们的工作首次证明了LLM能够自主识别其固有缺陷,并为未来的动态和自动评估提供了见解。

英文摘要:

The impressive performance of Large Language Models (LLMs) has consistently surpassed numerous human-designed benchmarks, presenting new challenges in assessing the shortcomings of LLMs. Designing tasks and finding LLMs' limitations are becoming increasingly important. In this paper, we investigate the question of whether an LLM can discover its own limitations from the errors it makes. To this end, we propose a Self-Challenge evaluation framework with human-in-the-loop. Starting from seed instances that GPT-4 fails to answer, we prompt GPT-4 to summarize error patterns that can be used to generate new instances and incorporate human feedback on them to refine these patterns for generating more challenging data, iteratively. We end up with 8 diverse patterns, such as text manipulation and questions with assumptions. We then build a benchmark, SC-G4, consisting of 1,835 instances generated by GPT-4 using these patterns, with human-annotated gold responses. The SC-G4 serves as a challenging benchmark that allows for a detailed assessment of LLMs' abilities. Our results show that only 44.96\% of instances in SC-G4 can be answered correctly by GPT-4. Interestingly, our pilot study indicates that these error patterns also challenge other LLMs, such as Claude-3 and Llama-3, and cannot be fully resolved through fine-tuning. Our work takes the first step to demonstrate that LLMs can autonomously identify their inherent flaws and provide insights for future dynamic and automatic evaluation.

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。

↑