List your company

Sycophancy Evaluation on NLP

49.25Sycophancy PreferenceSynthetic Data Intervention
47.272460.621273.9787.3188Jan 26, 2026

Evaluation results

MethodLinks
Synthetic Data InterventionBase Model=Gemma-2-2B2026.0149.25
Ours SAEBase Model=Gemma-2-2B, Probe Type=SAE2026.0150
Ours ResidBase Model=Gemma-2-2B, Probe Type=Residual2026.0150.32
Ours ResidModel=Gemma-2-9B, Probe Type=Residual2026.0179.88
Ours SAEModel=Gemma-2-9B, Probe Type=SAE2026.0183.36
Supervised Pinpoint TuningBase Model=Gemma-2-2B2026.0189.81
Untrained Gemma-2-2BBase Model=Gemma-2-2B2026.0191.26
Untrained Gemma-2-9BModel=Gemma-2-9B2026.0198.59
Synthetic Data InterventionModel=Gemma-2-9B2026.0198.6
Supervised Pinpoint TuningModel=Gemma-2-9B2026.0198.69