OptikalLLM-10K is a dataset containing 10,000 instruction–response pairs tailored for training or evaluating cybersecurity‑focused large language models. Each record pairs a user instruction with an expert response that conveys accurate, actionable and safe guidance. The dataset spans basic and advanced topics, includes adversarial prompts to improve model robustness, and covers compliance, incident response… See the full description on the dataset page: https://huggingface.co/datasets/Cogensec/OptikalLLM_10k_dataset.
