Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
CASTLEgithub.com · Low evidenceThe CASTLE Benchmark is a dataset of 250 hand-crafted C programs covering 25 common CWEs, designed to evaluate static code analyzers and LLMs for security vulnerability detection. It introduces the CASTLE Score metric, which balances true positive and false positive rates to enable fair comparisons of single tools and tool combinations.
Parse Score