Scientific Error Detection on SPOT Math/CS Equation and Proof Error Subset
89.7Detection AccuracyPAT
Evaluation Results
| Method | Links | |
|---|---|---|
| PATBase Model=Gemini 3.1 Pro2026.06 | 89.7 | |
| Gemini 3.1 Promode=Zero-Shot2026.06 | 55.2 | |
| Original SPOT SOTA2026.06 | 21.1 |
Loading the SOTA2 catalog…