Mathematical Reasoning on Average (GSM8K, MATH500, AMC23, AIME25)
72.5AccuracyBase
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BaseBackbone=DeepScaleR-1.5B-Preview2026.02 | 72.5 | 4,522 | 0 | |
| +RLOOBackbone=DeepScaleR-1.5B-Preview2026.02 | 72.5 | 4,182 | 0.08 | |
| DDCABackbone=DeepScaleR-1.5B-Preview2026.02 | 72.2 | 3,095 | 0.3 | |
| ThinkPrune-4kBackbone=DeepScaleR-1.5B-Preview2026.02 | 72.1 | 3,824 | 0.13 | |
| TLMREBackbone=DeepScaleR-1.5B-Preview2026.02 | 71 | 3,431 | 0.14 | |
| DDCABackbone=DeepSeek-R1-Distill-1.5B2026.02 | 68.5 | 4,908 | 0.42 | |
| +LPBackbone=DeepScaleR-1.5B-Preview2026.02 | 68.3 | 2,981 | 0.05 | |
| +RLOOBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 68.1 | 6,061 | 0.23 | |
| TLMREBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 66.8 | 6,205 | 0.14 | |
| ThinkPrune-4kBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 66.7 | 4,590 | 0.38 | |
| BaseBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 65 | 6,612 | 0 | |
| +LPBackbone=DeepSeek-R1-Distill-1.5B2026.02 | 64.6 | 4,747 | 0.25 |