arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2403.04307cs.CL
收藏

HaluEval-Wild:评估真实场景下语言模型的幻觉现象

HaluEval-Wild: Evaluating Hallucinations of Language Models in the Wild

Carnegie Mellon University(卡内基梅隆大学)

Zhiying Zhu, Yiming Yang, Zhiqing Sun

AI总结:

针对现有幻觉基准无法覆盖真实交互复杂性的问题,研究推出HaluEval-Wild基准,从ShareGPT筛选高难查询并分类,结合GPT-4与RAG生成答案,以细粒度评估真实场景下LLM的幻觉情况。

摘要:

幻觉问题对大语言模型(LLM)在关键领域的可靠性构成重大挑战。近期旨在评估传统NLP任务(如知识密集型问答(QA)和摘要生成)中LLM幻觉的基准,不足以捕捉动态真实场景下用户与LLM交互的复杂性。为填补这一空白,我们推出HaluEval-Wild,这是首个专门用于评估真实场景中LLM幻觉的基准。我们从现有真实世界用户-LLM交互数据集ShareGPT中,精心收集经Alpaca对抗性过滤的高难度用户查询,用以评估各类LLM的幻觉率。通过分析收集到的查询,我们将其划分为五种不同类型,从而可对LLM表现出的幻觉类型进行细粒度分析,并借助强大的GPT-4模型与检索增强生成(RAG)合成参考答案。我们的基准为加深对真实交互场景下LLM可靠性的理解并提升其可靠性提供了新途径。该基准可在https://github.com/HaluEval-Wild/HaluEval-Wild获取。

英文摘要:

Hallucinations pose a significant challenge to the reliability of large language models (LLMs) in critical domains. Recent benchmarks designed to assess LLM hallucinations within conventional NLP tasks, such as knowledge-intensive question answering (QA) and summarization, are insufficient for capturing the complexities of user-LLM interactions in dynamic, real-world settings. To address this gap, we introduce HaluEval-Wild, the first benchmark specifically designed to evaluate LLM hallucinations in the wild. We meticulously collect challenging (adversarially filtered by Alpaca) user queries from ShareGPT, an existing real-world user-LLM interaction datasets, to evaluate the hallucination rates of various LLMs. Upon analyzing the collected queries, we categorize them into five distinct types, which enables a fine-grained analysis of the types of hallucinations LLMs exhibit, and synthesize the reference answers with the powerful GPT-4 model and retrieval-augmented generation (RAG). Our benchmark offers a novel approach towards enhancing our comprehension of and improving LLM reliability in scenarios reflective of real-world interactions. Our benchmark is available at https://github.com/HaluEval-Wild/HaluEval-Wild.

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。

↑