LLM-generated Candidate Detection Rules: YAML outputs for 123 exploits

This dataset presents LLM-generated candidate detection rules in YAML format, produced by testing two commercial Large Language Models (LLMs): GPT-4o and Claude 3.5 Sonnet. The dataset is organized hierarchically, with directories for each model. The model directory contains six parameter variation folders named according to their 'temperature_top_p' configuration (e.g., 0.5_1, 0_0, 0_1, etc.), where the first value represents temperature and the second represents top_p. Each parameter folder contains YAML text files representing the model's output when tasked with generating candidate detection rules for 123 different exploits. The filenames correspond to exploit IDs from ExploitDB or PacketStorm, enabling direct reference to the original exploit samples.