OllaBench

该数据集采用了一个名为OllaBench的新颖评估框架，该框架用于评估大型语言模型（LLMs）在回答基于情景的信息安全合规性和非合规性问题时的一致性、准确性和冗余性。该数据集包含了21种不同类型问题的LLMs性能结果，这些问题涵盖了诸如“认知路径选择”、“角色识别”、“团队风险”和“目标因素”等。这项任务的目的是在网络安全情景中对大型语言模型进行评估。

This dataset utilizes a novel evaluation framework termed OllaBench, which is dedicated to evaluating the consistency, accuracy, and redundancy of large language models (LLMs) when responding to scenario-based questions concerning information security compliance and non-compliance. This dataset encompasses performance results of LLMs across 21 distinct question categories, covering topics such as Cognitive Path Selection, Role Recognition, Team Risk, and Target Factors, among others. The core objective of this task is to evaluate large language models within the context of cybersecurity scenarios.