Cybersecurity Reasoning + Tool Calling Dataset — 70,802 training entries in ChatML format. Includes both Q&A with reasoning and pentest tool call traces.

	📊 Dataset Overview

Split Entries Description

Train 70,802 63,216 Q&A + 7,586 pentest tool call traces (10.7%)

Valid 7,437 Q&A with cybersecurity reasoning

Test 3,719 Q&A with cybersecurity reasoning

Total 81,958 Combined

	📁 File Structure… See the full description on the dataset page: https://huggingface.co/datasets/hotdogs/uka-cyber-dataset.