触发LLM中的幻觉：大型语言模型中提示诱导幻觉的定量研究
Triggering Hallucinations in LLMs: A Quantitative Study of Prompt-Induced Hallucination in Large Language Models
AI总结：
本研究提出了一个基于提示的框架，通过幻觉诱导提示（HIP）和幻觉量化提示（HQP）系统地触发并量化大型语言模型中的幻觉，为研究幻觉脆弱性提供了可复现的测试平台。
摘要：
大型语言模型（LLMs）中的幻觉在从医疗到法律等现实应用中带来了日益严峻的挑战，在这些应用中事实的可靠性至关重要。尽管在 Alignment 和 Instruction Tuning 方面取得了进展，LLMs 仍可能生成流畅但根本不真实的输出。理解这些幻觉背后的认知动态仍是一个开放问题。在本研究中，我们提出了一个基于提示的框架来系统地触发和量化幻觉：一个是幻觉诱导提示（Hallucination-Inducing Prompt, HIP），它以误导的方式人工融合语义上遥远的概念（例如元素周期表和塔罗占卜）；另一个是幻觉量化提示（Hallucination Quantifying Prompt, HQP），它对输出的合理性、置信度和连贯性进行评分。跨多个 LLMs 的对照实验表明，与其 null-fusion 控制组相比，HIPs 持续产生连贯性更差且幻觉更多的响应。这些效应在不同模型间存在差异，推理导向的 LLMs 展现出与通用型模型截然不同的特征。我们的框架为研究幻觉脆弱性提供了可复现的测试平台，并为开发更安全、更具内省能力的 LLMs 打开了大门，使其能够检测并自我调节概念不稳定性的发作。
英文摘要：
Hallucinations in large language models (LLMs) present a growing challenge across real-world applications, from healthcare to law, where factual reliability is essential. Despite advances in alignment and instruction tuning, LLMs can still generate outputs that are fluent yet fundamentally untrue. Understanding the cognitive dynamics that underlie these hallucinations remains an open problem. In this study, we propose a prompt-based framework to systematically trigger and quantify hallucination: a Hallucination-Inducing Prompt (HIP), which synthetically fuses semantically distant concepts (e.g., periodic table of elements and tarot divination) in a misleading way, and a Hallucination Quantifying Prompt (HQP), which scores the plausibility, confidence, and coherence of the output. Controlled experiments across multiple LLMs revealed that HIPs consistently produced less coherent and more hallucinated responses than their null-fusion controls. These effects varied across models, with reasoning-oriented LLMs showing distinct profiles from general-purpose ones. Our framework provides a reproducible testbed for studying hallucination vulnerability, and opens the door to developing safer, more introspective LLMs that can detect and self-regulate the onset of conceptual instability.
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。
