Sycophancy Evaluation on Open-Ended Sycophancy
48.15Syc ScoreSynthetic Data Intervention
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Synthetic Data InterventionBase Model=Gemma-2-2B2026.01 | 48.15 | 50 | |
| Ours SAEModel=Gemma-2-9B, Probe Type=SAE2026.01 | 44.44 | 69.23 | |
| Ours ResidBase Model=Gemma-2-2B, Probe Type=Residual2026.01 | 44.44 | 53.85 | |
| Synthetic Data InterventionModel=Gemma-2-9B2026.01 | 40.74 | 46.15 | |
| Untrained Gemma-2-2BBase Model=Gemma-2-2B2026.01 | 37.04 | 69.23 | |
| Supervised Pinpoint TuningBase Model=Gemma-2-2B2026.01 | 37.04 | 69.23 | |
| Ours SAEBase Model=Gemma-2-2B, Probe Type=SAE2026.01 | 37.04 | 61.54 | |
| Untrained Gemma-2-9BModel=Gemma-2-9B2026.01 | 33.33 | 69.23 | |
| Supervised Pinpoint TuningModel=Gemma-2-9B2026.01 | 33.33 | 69.23 | |
| Ours ResidModel=Gemma-2-9B, Probe Type=Residual2026.01 | 29.63 | 69.23 |