PRIMUS
收藏资源简介:
PRIMUS数据集是由趋势科技AI实验室创建的一套开源数据集,包含预训练、指令微调和推理精炼三个阶段的数据集,旨在探索网络安全领域的大型语言模型训练。该数据集涵盖了多个网络安全任务,包括警报解释、可疑命令分析、安全事件查询生成等。数据集从多个来源收集高质量网络安全文本,并通过一系列预处理步骤,如语言模型过滤、去重、质量过滤等,形成了约2.58亿个标记的数据集。
The PRIMUS dataset is an open-source dataset created by Trend Micro AI Lab. It includes three stages of datasets for pre-training, instruction fine-tuning, and reasoning refinement, with the goal of exploring large language model (LLM) training in the cybersecurity domain. This dataset covers multiple cybersecurity tasks, including alert interpretation, suspicious command analysis, security event query generation, and others. It collects high-quality cybersecurity text from various sources, and undergoes a series of preprocessing steps such as language model filtering, deduplication, and quality filtering, resulting in a dataset with approximately 258 million tokens.

- 1Primus: A Pioneering Collection of Open-Source Datasets for Cybersecurity LLM Training趋势科技AI实验室 · 2025年



