Adversarial Detection on Expanded benchmark Arithmetic error
19Caught RateTheoria
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TheoriaModel=GPT-5.5, Web search=Enabled, Evaluation protocol=Structured (step-level typed judges with BEFORE→AFTER diffs)2026.07 | 19 | 95 | |
| Holistic judgeModel=GPT-5.5, Web search=Enabled, Evaluation protocol=Holistic (single-call global assessment)2026.07 | 19 | 95 |