ResearchBenchmarksVulnerability Reproduction on FL-Bench 100 (standardized)Follow15Success Rate (SR)PoC-Adapt11.8812.6913.514.31Apr 8, 2026Evaluation ResultsMethodMethodLinksSuccess Rate (SR)Time To Reproduce (TTE) (steps)Exploitation Effectiveness (EE)PoC-AdaptInference budget=$1, T...Inference budget=$1, Timeout=60-minute, Refinement loops=32026.041516.332.5FaultLineInference budget=$1, T...Inference budget=$1, Timeout=60-minute, Refinement loops=32026.041235.921.1