Mathematical Reasoning on MATH 500 (Token Usage Metrics)
2,025,345Avg Token Usage (R0)Elim
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ElimModel=GPT-4.1-mini, Seeds=42026.04 | 2,025,345 | 1,680,173 | 1,308,144 | 1,249,419 | 6,263,081 | |
| Self-Evolving Test-Time AllocationModel=GPT-4.1-mini, Seeds=42026.04 | 2,025,345 | 1,102,486 | 902,109 | 741,611 | 4,771,551 |