List your company

Sycophancy Mitigation on Sycophancy

36.2SycophancySFT (NO RLHF)
34.75244.52654.364.074Feb 2, 2026

Evaluation results

MethodLinks
SFT (NO RLHF)RLHF=None2026.0236.241.3
ARAFramework=Auditor-gated RLHF2026.0238.477.2
INFORMCategory=Regularization Methods2026.0247.262.1
ODINCategory=Regularization Methods2026.0251.663.4
RM ENSEMBLECategory=Reward Model Improvements2026.0252.864.6
FILTERING (R > mu + 2sigma)Category=Training Interventions2026.0255.150.3
PPO W/KLCategory=Regularization Methods2026.0258.368.2
PPOCondition=Unmitigated2026.0272.476.8