Pairwise anchor type classification on Sycophantic Anchors 1.0 (test)
84.6Syco vs Correct AccuracyR1-Distill-Llama-8B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| R1-Distill-Llama-8BLayer=282026.01 | 84.6 | 77.5 | 64 | 13.5 | |
| Falcon-H1R-7BLayer=342026.01 | 79.3 | 75.5 | 72.2 | 3.3 | |
| R1-Distill-Qwen-7BLayer=212026.01 | 76.1 | 73.2 | 70.1 | 3.1 | |
| R1-Distill-Qwen-1.5BLayer=212026.01 | 73.8 | 76.9 | 70.6 | 6.3 |