Sycophancy Evaluation on NLP
49.25Sycophancy PreferenceSynthetic Data Intervention
Evaluation Results
| Method | Links | |
|---|---|---|
| Synthetic Data InterventionBase Model=Gemma-2-2B2026.01 | 49.25 | |
| Ours SAEBase Model=Gemma-2-2B, Probe Type=SAE2026.01 | 50 | |
| Ours ResidBase Model=Gemma-2-2B, Probe Type=Residual2026.01 | 50.32 | |
| Ours ResidModel=Gemma-2-9B, Probe Type=Residual2026.01 | 79.88 | |
| Ours SAEModel=Gemma-2-9B, Probe Type=SAE2026.01 | 83.36 | |
| Supervised Pinpoint TuningBase Model=Gemma-2-2B2026.01 | 89.81 | |
| Untrained Gemma-2-2BBase Model=Gemma-2-2B2026.01 | 91.26 | |
| Untrained Gemma-2-9BModel=Gemma-2-9B2026.01 | 98.59 | |
| Synthetic Data InterventionModel=Gemma-2-9B2026.01 | 98.6 | |
| Supervised Pinpoint TuningModel=Gemma-2-9B2026.01 | 98.69 |