arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2501.03491cs.CLcs.AI
收藏

大语言模型能提出好问题吗?

Can LLMs Ask Good Questions?

University of California Berkeley(加州大学伯克利分校);KACST(王国立科学与技术委员会);University of Washington(华盛顿大学)

Yueheng Zhang, Xiaoyuan Liu, Yiyou Sun, Atheer Alharbi, Hend Alzahrani, Tianneng Shi, Basel Alomair, Dawn Song

AI总结:

本研究从六个维度比较了LLM与人类生成的问题,发现LLM问题更均匀覆盖上下文且要求更长答案,为问题质量研究提供参考。

摘要:

我们评估了由大语言模型(LLMs)根据上下文生成的问题,并将其与人类撰写的问题在六个维度上进行比较:问题类型、问题长度、上下文覆盖度、可回答性、非普遍性以及所需答案长度。我们的研究涵盖了两种开源和两种专有的最先进模型。结果表明,与问答任务中常见的位置偏差相比,LLM生成的问题往往要求更长的描述性答案,并表现出更均匀分布的上下文关注点。这些发现为LLM生成问题的独特特征提供了见解,并为未来关于问题质量和下游应用的研究提供了参考。

英文摘要:

We evaluate questions generated by large language models (LLMs) from context, comparing them to human-authored questions across six dimensions: question type, question length, context coverage, answerability, uncommonness, and required answer length. Our study spans two open-source and two proprietary state-of-the-art models. Results reveal that LLM-generated questions tend to demand longer descriptive answers and exhibit more evenly distributed context focus, in contrast to the positional bias often seen in QA tasks. These findings provide insights into the distinctive characteristics of LLM-generated questions and inform future work on question quality and downstream applications.

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。

↑