ResearchBenchmarksOver-refusal Compliance on XS (test)Follow98Compliance Rate (Keyword Filter)DCR7278.7585.592.25Feb 10, 2026Evaluation ResultsMethodMethodLinksCompliance Rate (Keyword Filter)Compliance Rate (GPT-4o Judge)Compliance Rate (GPT 5.1 Judge)DCR2026.02989796SCANS2026.02838284Surgical2026.02817379STL-aug2026.02757572STL2026.02737472