List your company

Sycophancy Evaluation on Open-Ended Sycophancy

48.15Syc ScoreSynthetic Data Intervention
28.889233.889638.8943.8904Jan 26, 2026

Evaluation results

MethodLinks
Synthetic Data InterventionBase Model=Gemma-2-2B2026.0148.1550
Ours SAEModel=Gemma-2-9B, Probe Type=SAE2026.0144.4469.23
Ours ResidBase Model=Gemma-2-2B, Probe Type=Residual2026.0144.4453.85
Synthetic Data InterventionModel=Gemma-2-9B2026.0140.7446.15
Untrained Gemma-2-2BBase Model=Gemma-2-2B2026.0137.0469.23
Supervised Pinpoint TuningBase Model=Gemma-2-2B2026.0137.0469.23
Ours SAEBase Model=Gemma-2-2B, Probe Type=SAE2026.0137.0461.54
Untrained Gemma-2-9BModel=Gemma-2-9B2026.0133.3369.23
Supervised Pinpoint TuningModel=Gemma-2-9B2026.0133.3369.23
Ours ResidModel=Gemma-2-9B, Probe Type=Residual2026.0129.6369.23