HalluciNot：通过上下文与常识验证进行幻觉检测
HalluciNot: Hallucination Detection Through Context and Common Knowledge Verification
AI总结：
本文提出企业级LLM幻觉检测系统HalluciNot及模型HDM-2，通过上下文与常识双重验证并提供词级标注，在多个数据集上优于现有方法。
摘要：
本文介绍了一个用于在企业环境中检测大语言模型（LLM）输出幻觉的综合系统。我们提出了一种针对企业应用中幻觉的LLM响应新分类法，将其分为基于上下文的、常识性的、企业特有的和无害的陈述。我们的幻觉检测模型HDM-2在上下文和普遍已知事实（常识）两方面对LLM响应进行验证。它同时提供幻觉分数和词级标注，从而能够精确识别有问题的内容。为了在基于上下文和常识性幻觉上对其进行评估，我们引入了一个新数据集HDMBench。实验结果表明，HDM-2在RagTruth、TruthfulQA和HDMBench数据集上均优于现有方法。这项工作解决了企业部署中的具体挑战，包括计算效率、领域专业化和细粒度错误识别。我们的评估数据集、模型权重和推理代码均已公开。
英文摘要：
This paper introduces a comprehensive system for detecting hallucinations in large language model (LLM) outputs in enterprise settings. We present a novel taxonomy of LLM responses specific to hallucination in enterprise applications, categorizing them into context-based, common knowledge, enterprise-specific, and innocuous statements. Our hallucination detection model HDM-2 validates LLM responses with respect to both context and generally known facts (common knowledge). It provides both hallucination scores and word-level annotations, enabling precise identification of problematic content. To evaluate it on context-based and common-knowledge hallucinations, we introduce a new dataset HDMBench. Experimental results demonstrate that HDM-2 out-performs existing approaches across RagTruth, TruthfulQA, and HDMBench datasets. This work addresses the specific challenges of enterprise deployment, including computational efficiency, domain specialization, and fine-grained error identification. Our evaluation dataset, model weights, and inference code are publicly available.
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。
