HDMBENCH
收藏资源简介:
HDMBENCH是一个新的数据集,由AIMon Labs创建,用于评估语言模型在企业和常见知识背景下的虚构检测。该数据集包含针对大型语言模型输出的虚构响应,特别是基于上下文和常见知识的虚构。它旨在解决企业环境中对准确性和可靠性的高要求,以及虚构内容可能引起的法律风险、合规问题和对用户信任的侵蚀。数据集的具体内容、创建过程和应用领域在论文中未详细描述。
HDMBENCH is a novel dataset created by AIMon Labs for evaluating hallucination detection in large language models (LLMs) under enterprise and general knowledge contexts. This dataset includes hallucinatory responses within the outputs of large language models, specifically those based on contextual information and general knowledge. It aims to address the high demands for accuracy and reliability in enterprise environments, as well as the legal risks, compliance issues, and erosion of user trust that hallucinatory content may induce. The specific content, creation process, and application domains of the dataset are not described in detail in the paper.

- 1HalluciNot: Hallucination Detection Through Context and Common Knowledge VerificationAIMon Labs · 2025年



