arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2409.05746stat.MLcs.LG
收藏

LLM 将永远产生幻觉,而我们必须接受这一点

LLMs Will Always Hallucinate, and We Need to Live With This

DataLabs(数据实验室);United We Care

Sourav Banerjee, Ayushi Agarwal, Saloni Singla

AI总结:

本文基于计算理论与哥德尔第一不完备定理论证 LLM 幻觉是其数学逻辑结构导致的必然现象,并提出结构性幻觉概念,否定完全消除幻觉的可能。

摘要:

随着大语言模型在各个领域变得无处不在,批判性地审视其固有局限变得十分重要。本工作认为,语言模型中的幻觉并非偶发错误,而是这些系统不可避免的特征。我们证明,幻觉源于 LLM 的基础数学与逻辑结构。因此,不可能通过架构改进、数据集增强或事实核查机制来消除幻觉。我们的分析借鉴了计算理论和哥德尔第一不完备定理,该定理涉及停机问题、空集问题和接受问题等问题的不可判定性。我们证明,LLLM 流程的每个阶段——从训练数据汇编到事实检索、意图分类和文本生成——产生幻觉的概率都不为零。本工作提出“结构性幻觉”概念,将其作为这些系统的内在本质。通过确立幻觉在数学上的必然性,我们挑战了幻觉可以被完全缓解这一主流观点。

英文摘要:

As Large Language Models become more ubiquitous across domains, it becomes important to examine their inherent limitations critically. This work argues that hallucinations in language models are not just occasional errors but an inevitable feature of these systems. We demonstrate that hallucinations stem from the fundamental mathematical and logical structure of LLMs. It is, therefore, impossible to eliminate them through architectural improvements, dataset enhancements, or fact-checking mechanisms. Our analysis draws on computational theory and Godel's First Incompleteness Theorem, which references the undecidability of problems like the Halting, Emptiness, and Acceptance Problems. We demonstrate that every stage of the LLM process-from training data compilation to fact retrieval, intent classification, and text generation-will have a non-zero probability of producing hallucinations. This work introduces the concept of Structural Hallucination as an intrinsic nature of these systems. By establishing the mathematical certainty of hallucinations, we challenge the prevailing notion that they can be fully mitigated.

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。

↑