Training model: o3
Inference model: gpt-4o-mini
| Dataset | Validation accuracy | Test accuracy |
|---|---|---|
| espionage | 0.95 | 0.8 |
| potions | 0.8 | 0.75 |
| southgermancredit | 0.5961538461538461 | 0.49557522123893805 |
| timetravel_insurance | 0.8 | 0.7 |
| titanic | 0.803921568627451 | 0.6862745098039216 |
| wisconsin | 0.9365079365079365 | 0.8333333333333334 |