Mathematical Reasoning on AIME 2025 (Latency and Budget)
4.95Latency (ms/tok)DDTree-oracle
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DDTree-oracleBackbone=Qwen3-4B, Temperature=0.02026.06 | 4.95 | 9.4 | 128 | |
| CADDTREEBackbone=Qwen3-4B, Temperature=0.02026.06 | 4.96 | 9.64 | 215 | |
| CADDTREEBackbone=Qwen3-8B, Temperature=0.02026.06 | 5 | 9.51 | 156 | |
| DDTree-oracleBackbone=Qwen3-8B, Temperature=0.02026.06 | 5.13 | 9.18 | 128 | |
| DFlashBackbone=Qwen3-8B, Temperature=0.02026.06 | 6.11 | 7.49 | — | |
| DFlashBackbone=Qwen3-4B, Temperature=0.02026.06 | 6.28 | 7.43 | — | |
| ARBackbone=Qwen3-8B, Temperature=0.02026.06 | 31.44 | — | — | |
| ARBackbone=Qwen3-4B, Temperature=0.02026.06 | 31.64 | — | — |