Mathematical Reasoning on AIME 2010-2023
100CoverageDASE overall
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DASE overallN=254, Cost per query=$0.041/q‡2026.05 | 100 | 85.8 | — | |
| Opus overallN=254, Cost per query=$0.061/q†2026.05 | 100 | 84.7 | — | |
| Opus (conf≥97)Bucket=High conf., N=132, Confidence Threshold=conf≥972026.05 | 52 | 97 | 25.7 | |
| DASE W=2 (3-seed)Bucket=Left-wall, N=1292026.05 | 50.8 | 73.6 | — | |
| DASE W=2 (3-seed)Bucket=Right-wall, N=1252026.05 | 49.2 | 97.1 | 24.8 | |
| Opus (conf<97)Bucket=Low conf., N=122, Confidence Threshold=conf<972026.05 | 48 | 71.3 | — |