Mathematical Reasoning on AIME 24 (Acc, Len, TE)
57.4AccuracySTACK
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| STACKBackbone=DeepSeek-R1-Distill-Qwen-7B2026.04 | 57.4 | 7,274 | 0.79 | |
| TokenSqueezeBackbone=DeepSeek-R1-Distill-Qwen-7B2026.04 | 56.3 | 9,079 | 0.62 | |
| MuTISBackbone=DeepSeek-R1-Distill-Qwen-7B2026.04 | 54.7 | 6,548 | 0.84 | |
| OriginalBackbone=DeepSeek-R1-Distill-Qwen-7B2026.04 | 53.8 | 13,178 | 0.41 | |
| PromptBackbone=DeepSeek-R1-Distill-Qwen-7B2026.04 | 53.4 | 11,462 | 0.47 | |
| ConCISEBackbone=DeepSeek-R1-Distill-Qwen-7B2026.04 | 48.5 | 7,519 | 0.64 | |
| STACKBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 36.7 | 8,752 | 0.42 | |
| TokenSqueezeBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 33.3 | 10,672 | 0.31 | |
| MuTISBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 31.2 | 8,976 | 0.35 | |
| ConCISEBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 30.1 | 9,315 | 0.32 | |
| OriginalBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 29.4 | 16,814 | 0.17 | |
| PromptBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 28.3 | 15,892 | 0.18 |