joduor/adaption-biosecurity-risk-labels

该数据集包含与生物技术、基因工程和实验室安全相关的提示对，每个提示对都标有风险评估。完成部分将输入分类为LOW_RISK（用于一般教育或安全主题）或HIGH_RISK_INFOHAZARD（用于详细描述有害程序，如病原体合成或绕过防护措施的请求）。它作为评估模型对生物信息危害的安全过滤能力的基准。

This dataset contains pairs of prompts related to biotechnology, genetic engineering, and laboratory safety, each labeled with a risk assessment. The completions classify inputs as either LOW_RISK for general educational or safety topics, or HIGH_RISK_INFOHAZARD for requests detailing harmful procedures like pathogen synthesis or containment bypass. It serves as a benchmark for evaluating model safety filters against biological information hazards.