| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| Anthropic Model-Written Evals | BCT | Sycophancy Rate0.78 | 45 | 1mo ago | |
| Held-out | BCT | BRR43.6 | 40 | 1mo ago | |
| MMLU on-the-fly | BCT | BRR0.524 | 40 | 1mo ago | |
| TruthfulQA | Linear Probe MHA | Sycophancy Rate25 | 14 | 4mo ago | |
| Sycophancy | SFT (NO RLHF) | Sycophancy36.2 | 8 | 4mo ago | |
| BCT Sycophancy (held-out) | BRR Ratio100 | 4 | 1mo ago |