Mathematical Reasoning on AMC23 (Acc., #Tok.)
83.2AccuracyDDCA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DDCABackbone=DeepScaleR-1.5B-Preview2026.02 | 83.2 | 3,515 | |
| +RLOOBackbone=DeepScaleR-1.5B-Preview2026.02 | 82.3 | 4,468 | |
| BaseBackbone=DeepScaleR-1.5B-Preview2026.02 | 81.7 | 4,993 | |
| ThinkPrune-4kBackbone=DeepScaleR-1.5B-Preview2026.02 | 81.3 | 4,179 | |
| +LPBackbone=DeepScaleR-1.5B-Preview2026.02 | 79.4 | 3,226 | |
| TLMREBackbone=DeepScaleR-1.5B-Preview2026.02 | 79.1 | 3,734 | |
| DDCABackbone=DeepSeek-R1-Distill-1.5B2026.02 | 76.3 | 5,556 | |
| +RLOOBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 76.1 | 6,953 | |
| +LPBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 74.5 | 5,336 | |
| ThinkPrune-4kBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 72.8 | 5,107 | |
| TLMREBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 72.2 | 7,319 | |
| BaseBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 71.8 | 7,945 |