Training model: claude-sonnet-4-20250514
Inference model: gpt-4.1-mini
| Dataset | Validation accuracy | Test accuracy |
|---|---|---|
| espionage | 0.9 | 0.95 |
| potions | 0.7 | 0.45 |
| southgermancredit | 0.6923076923076923 | 0.7256637168141593 |
| timetravel_insurance | 0.75 | 0.7 |
| titanic | 0.8431372549019608 | 0.7254901960784313 |
| wisconsin | 0.873015873015873 | 0.8484848484848485 |
| espionage | 0.9 | 0.95 |
| potions | 0.75 | 0.7 |
| timetravel_insurance | 0.75 | 0.75 |
| espionage | 0.95 | 1.0 |
| potions | 0.75 | 0.6 |
| timetravel_insurance | 0.0 | 0.0 |
| wisconsin | 0.746031746031746 | 0.803030303030303 |