List your company

Sycophancy Evaluation on PHIL

99.34Sycophancy PreferenceSupervised Pinpoint Tuning
48.182461.463774.74588.0263Jan 26, 2026

Evaluation results

MethodLinks
Supervised Pinpoint TuningModel=Gemma-2-9B2026.0199.34
Synthetic Data InterventionModel=Gemma-2-9B2026.0198.73
Untrained Gemma-2-9BModel=Gemma-2-9B2026.0198.71
Supervised Pinpoint TuningBase Model=Gemma-2-2B2026.0190.41
Untrained Gemma-2-2BBase Model=Gemma-2-2B2026.0190.35
Synthetic Data InterventionBase Model=Gemma-2-2B2026.0179.65
Ours ResidModel=Gemma-2-9B, Probe Type=Residual2026.0169.56
Ours SAEModel=Gemma-2-9B, Probe Type=SAE2026.0160.81
Ours ResidBase Model=Gemma-2-2B, Probe Type=Residual2026.0153.98
Ours SAEBase Model=Gemma-2-2B, Probe Type=SAE2026.0150.15