URL extraction results: testing on-premise LLMs on selected exploits

This dataset presents URL extraction results from testing three on-premise Large Language Models (LLMs): Llama 3.1: 8B, Qwen 2.5 Coder: 7B Instruct, and Codestral: 22B. Each model has its directory containing subdirectories organized by CVE names, which in turn contain text files named after corresponding exploit IDs from ExploitDB or PacketStorm. These text files represent each model's output when tasked with extracting URLs that would be accessed during the execution of the exploit code, enabling direct reference to the original exploit samples.