arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2512.11509cs.CLcs.AI

减少幻觉是否意味着减少创造力？在大语言模型中的实证研究

Does Less Hallucination Mean Less Creativity? An Empirical Investigation in LLMs

Mohor Banerjee, Nadya Yuki Wangsajaya, Syed Ali Redha Alsagoff, Min Sen Tan, Zachary Choy Kit Chun, Alvin Chan Guo Wei

备注：
Accepted at the AAAI 2026 Workshop on AI for Scientific Research (AI4Research)
阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

本研究探讨了三种减少幻觉的方法对大语言模型创造力的影响，发现CoVe促进发散思维，DoLa抑制创造力，RAG影响较小，为科学应用中的准确性与创造性平衡提供指导。

摘要：

大语言模型（LLMs）在自然语言理解和推理方面表现出卓越的能力，但存在幻觉问题：生成事实性错误内容。尽管已开发出许多减少幻觉的方法，但这些方法对创造性生成的影响尚未被探索。这一空白在需要事实准确性和创造性假设生成的AI辅助科学发现中尤为重要。我们研究了三种减少幻觉的技术：验证链（CoVe）、层间对比解码（DoLa）和检索增强生成（RAG），这些技术如何影响LLMs的创造力。在多个模型家族（LLaMA、Qwen、Mistral）上，针对两个创造力基准（NeoCoder和CS4）进行评估，我们发现这些方法对发散创造力有相反的影响。CoVe增强发散思维，DoLa抑制它，而RAG影响最小。我们的发现为在科学应用中选择适当的幻觉减少方法提供了指导，其中事实准确性与创造性探索的平衡至关重要。

英文摘要：

Large Language Models (LLMs) exhibit remarkable capabilities in natural language understanding and reasoning, but suffer from hallucination: the generation of factually incorrect content. While numerous methods have been developed to reduce hallucinations, their impact on creative generations remains unexplored. This gap is particularly critical for AI-assisted scientific discovery, which requires both factual accuracy and creative hypothesis generation. We investigate how three hallucination-reduction techniques: Chain of Verification (CoVe), Decoding by Contrasting Layers (DoLa), and Retrieval-Augmented Generation (RAG), affect creativity in LLMs. Evaluating multiple model families (LLaMA, Qwen, Mistral) at varying scales (1B - 70B parameters) on two creativity benchmarks (NeoCoder and CS4), we find that these methods have opposing effects on divergent creativity. CoVe enhances divergent thinking, DoLa suppresses it, and RAG shows minimal impact. Our findings provide guidance for selecting appropriate hallucination-reduction methods in scientific applications, where the balance between factual accuracy and creative exploration is crucial.

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。

↑