Mathematical Reasoning on AIME 24 (Sel-AUC, AUROC)
92.5Sel-AUCStructU+Self-ConsU
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| StructU+Self-ConsUModel=DeepSeek R1, Aggregation=across2026.06 | 92.5 | 100 | |
| SemanticUModel=Amazon Nova Premier2026.06 | 29.4 | — | |
| StructUModel=Qwen 3 32B, Aggregation=within2026.06 | 23.5 | — |