ResearchBenchmarksSycophancy Evaluation on Sycophancy Evaluation Suite All Guardrail ConditionsFollow1.83Mean Sycophancy ScoreGen 2.51.4141.5221.631.738Apr 19, 2026Evaluation ResultsMethodMethodLinksMean Sycophancy Score95% Confidence IntervalSample Size (N)Gen 2.5Generation=2.5Generation=2.52026.041.831.793,225Gen 3.0Generation=3.0Generation=3.02026.041.481.453,265Gen 2.0Generation=2.0Generation=2.02026.041.431.42,340