Primus-Reasoning: Cybersecurity Tasks for LLM Training

PRIMUS is a pioneering collection of open-source datasets for cybersecurity LLM training. The Primus-Reasoning subset contains multiple cybersecurity reasoning tasks sourced from CTI-Bench, including CTI-RCM, CTI-VSP, CTI-ATE, and CTI-MCQ. It was augmented in June 2025 with distilled samples from DeepSeek-R1, incorporating intermediate reasoning steps and final answers.

Use Cases

• Training LLMs on cybersecurity reasoning tasks based on the CTI-Bench framework mentioned in the description
• Fine-tuning models for cyber threat intelligence analysis based on the dataset's stated purpose
• Benchmarking LLM performance on structured cybersecurity questions based on the included task types
• Studying the impact of incorporating intermediate reasoning steps from models like DeepSeek-R1 on training outcomes

Strengths

• Dataset was augmented with distilled samples from DeepSeek-R1 on 2025-06-02, indicating recent updates
• Focuses on multiple specific cybersecurity reasoning tasks (CTI-RCM, CTI-VSP, CTI-ATE, CTI-MCQ) as stated
• Includes both intermediate reasoning steps and final answers, as described for the DeepSeek-R1 samples

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Data may reflect bias inherent to the specific CTI-Bench sources and the distillation process from DeepSeek-R1

Provenance

Source

trendmicro-ailab

Collection Method

Likely compiled from CTI-Bench tasks and augmented with distilled samples from DeepSeek-R1.

Freshness

Last updated 2025-06-02 11:27:07