ResearchBenchmarksBlack-box system prompt extraction on 41 Models Open-weightFollow2.4Avg TurnsJUSTASK2.282.342.42.46Jan 29, 2026Evaluation ResultsMethodMethodLinksAvg TurnsSuccess RateJUSTASKModels=23, Primary Ski...Models=23, Primary Skills=L14, L7, H42026.012.4100