在大型语言模型中诱导知识谦逊：一种针对减少幻觉的定向SFT方法
Inducing Epistemological Humility in Large Language Models: A Targeted SFT Approach to Reducing Hallucination
AI总结：
本文提出HypoTermInstruct数据集，通过非现实术语问题训练模型知识谦逊，减少幻觉，实验显示其显著提升HypoTerm Score和FactScore，同时保持MMLU性能稳定。
摘要：
大型语言模型（LLMs）常出现幻觉，产生流畅但虚假的信息，部分原因在于监督微调（SFT）隐式奖励总是回应。我们引入HypoTermInstruct，一个SFT数据集（31,487个响应对应11,151个问题），旨在教导模型知识谦逊——识别自身知识限制并承认不确定性的能力。这通过关于不存在的“假设”术语的问题实现。我们还发布了HypoTermQA-Enhanced，一个通过多重验证强化幻觉倾向的基准。我们进行了800次受控LoRA SFT运行，覆盖Llama3.1-8B和Gemma3-4B（基础和指导版本），测试100种微调配置及其配对对照。结果表明，用HypoTermInstruct替代通用指令数据显著提高HypoTerm Score（中位数增加0.19%至25.91%）和FactScore（+0.39%至+0.86%），同时保持MMLU性能稳定（最小下降0.26%至0.35%）。我们的工作表明，针对性、高质量的SFT数据教导元认知技能可以有效减少幻觉，无需偏好/RL流程，提供机理洞察和可靠AI系统的实用路径。
英文摘要：
Large language models (LLMs) often hallucinate, producing fluent but false information, partly because supervised fine-tuning (SFT) implicitly rewards always responding. We introduce $\textit{HypoTermInstruct}$, an SFT dataset (31,487 responses for 11,151 questions) designed to teach models epistemological humility-the ability to recognize the limits of their own knowledge and admit uncertainty. This is achieved through questions about non-existent "hypothetical" terms. We also release $\textit{HypoTermQA-Enhanced}$, a benchmark for hallucination tendency strengthened through multiple validations. We conducted 800 controlled LoRA SFT runs across $\textit{Llama3.1-8B}$ and $\textit{Gemma3-4B}$ (base and instruct), testing 100 fine-tuning configurations with paired controls. Our results demonstrate that replacing generic instruction data with $\textit{HypoTermInstruct}$ significantly improves the HypoTerm Score (median increases of 0.19% to 25.91%) and FactScore (+0.39% to +0.86%), while maintaining stable performance on MMLU (minimal decreases of 0.26% to 0.35%). Our work demonstrates that targeted, high-quality SFT data teaching meta-cognitive skills can effectively reduce hallucination without preference/RL pipelines, providing mechanistic insights and a practical path toward more reliable AI systems.
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。
