arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2311.09702cs.CLcs.AI
收藏

推理链上的欺骗性语义捷径:模型在不产生幻觉的情况下能走多远?

Deceptive Semantic Shortcuts on Reasoning Chains: How Far Can Models Go without Hallucination?

University of California, Davis(加利福尼亚大学戴维斯分校);University of Pennsylvania(宾夕法尼亚大学);University of Southern California(南加利福尼亚大学)

Bangzheng Li, Ben Zhou, Fei Wang, Xingyu Fu, Dan Roth, Muhao Chen

AI总结:

本文提出EureQA探测基准,通过递归遮蔽实体并植入语义干扰线索,评估LLM能否沿证据链推理而非依赖关键词偏误走捷径。

摘要:

尽管大语言模型(LLM)近期取得进展,并在众多基准测试中表现出色,但最近研究揭示,LLM存在幻觉和不忠实推理问题。本文研究由语义关联诱发的一类特定幻觉。具体而言,我们考察LLM在多大程度上利用提示中的某些关键词或实体偏误走捷径,而不是遵循正确的推理路径。为量化这一现象,我们提出一种名为EureQA的新型探测方法与基准。我们从LLM能够以最高确定性答对的问题出发,利用证据句递归遮蔽重要实体,要求模型在回答问题前,根据证据链找出被遮蔽的实体。在证据构建过程中,我们有意将可能导向正确答案的语义线索(实体)替换为干扰线索(证据);这些线索不会直接导向正确答案,而是需要链式推理过程。我们评估模型能否遵循正确推理链,而不是通过干扰线索走捷径。我们发现,现有LLM缺乏遵循正确推理路径并抵抗贪婪捷径尝试所需的能力。研究表明,干扰性语义关联常导致模型幻觉,这为质疑当前LLM推理的有效性提供了有力证据。

英文摘要:

Despite the recent advancement in large language models (LLMs) and their high performances across numerous benchmarks, recent research has unveiled that LLMs suffer from hallucinations and unfaithful reasoning. This work studies a specific type of hallucination induced by semantic associations. Specifically, we investigate to what extent LLMs take shortcuts from certain keyword/entity biases in the prompt instead of following the correct reasoning path. To quantify this phenomenon, we propose a novel probing method and benchmark called EureQA. We start from questions that LLMs will answer correctly with utmost certainty, and mask the important entity with evidence sentence recursively, asking models to find masked entities according to a chain of evidence before answering the question. During the construction of the evidence, we purposefully replace semantic clues (entities) that may lead to the correct answer with distractor clues (evidence) that will not directly lead to the correct answer but require a chain-like reasoning process. We evaluate if models can follow the correct reasoning chain instead of short-cutting through distractor clues. We find that existing LLMs lack the necessary capabilities to follow correct reasoning paths and resist the attempt of greedy shortcuts. We show that the distractor semantic associations often lead to model hallucination, which is strong evidence that questions the validity of current LLM 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。

↑