Mathematical Reasoning on SVAMP (Acc, SL, SE)
96.7AccuracyL1
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| L12026.03 | 96.7 | 1,138 | 8.5 | |
| Cut1024max tokens=10242026.03 | 96.67 | 1,872 | 5.16 | |
| ROM_CSCCounterfactual Self-Correction=true2026.03 | 95.8 | 906 | 10.57 | |
| Vanilla2026.03 | 95 | 1,874 | 5.07 | |
| EAT2026.03 | 95 | 1,480 | 6.42 | |
| Cut2048max tokens=20482026.03 | 95 | 1,872 | 5.07 | |
| ROM2026.03 | 92.5 | 1,241 | 7.45 | |
| Cut512max tokens=5122026.03 | 86.67 | 1,872 | 4.63 | |
| CESBackbone=DeepSeek-R1-Distill-1.5B2026.05 | 86.4 | 620 | — | |
| DAPOBackbone=DeepSeek-R1-Distill-1.5B2026.05 | 85.6 | 618 | — |