Mathematical Reasoning on MATH-500 (Accuracy, Token Count, Calibration Rate)
93.5AccuracyVanilla
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VanillaBackbone=Ministral-3-8B-Reasoning-25122026.03 | 93.5 | 6,212 | 100 | |
| VanillaBackbone=Ministral-3-14B-Reasoning-25122026.03 | 93 | 6,385 | 100 | |
| OriginalBackbone=DeepSeek-R1-Distill-Qwen-7B2025.08 | 92.2 | 4,223 | — | |
| VanillaBackbone=Qwen3-14B2026.03 | 92 | 4,598 | 100 | |
| LCPOBackbone=DeepSeek-R1-Distill-Qwen-7B2025.08 | 91.4 | 2,033 | — | |
| DASTBackbone=DeepSeek-R1-Distill-Qwen-7B2025.08 | 91.2 | 3,563 | — | |
| VanillaBackbone=Qwen3-8B2026.03 | 91.1 | 5,037 | 100 | |
| TERMINATORBackbone=Qwen3-8B2026.03 | 90.7 | 2,425 | 45.1 | |
| TERMINATORBackbone=Qwen3-14B2026.03 | 90.7 | 2,261 | 46.8 | |
| TERMINATORBackbone=Ministral-3-14B-Reasoning-25122026.03 | 90.2 | 2,946 | 43.9 | |
| TrEffBackbone=DeepSeek-R1-Distill-Qwen-7B2025.08 | 90.2 | 2,413 | — | |
| Thought-CalibBackbone=Qwen3-8B2026.03 | 90.1 | 4,372 | 93.9 | |
| CoDBackbone=DeepSeek-R1-Distill-Qwen-7B2025.08 | 90.06 | 2,778 | — | |
| Thought-CalibBackbone=Qwen3-14B2026.03 | 89.8 | 3,778 | 92 | |
| L1-ExactBackbone=DeepSeek-R1-Distill-Qwen-7B2025.08 | 89.8 | 3,555 | — | |
| TERMINATORBackbone=Ministral-3-8B-Reasoning-25122026.03 | 89.1 | 2,863 | 47.8 | |
| L1-MaxBackbone=DeepSeek-R1-Distill-Qwen-7B2025.08 | 88.8 | 2,016 | — | |
| DynasorBackbone=Ministral-3-8B-Reasoning-25122026.03 | 88.1 | 2,967 | 56.8 | |
| Thought-CalibBackbone=Ministral-3-8B-Reasoning-25122026.03 | 87.7 | 5,695 | 87.8 | |
| Thought-CalibBackbone=Ministral-3-14B-Reasoning-25122026.03 | 87.3 | 5,860 | 95.9 | |
| DynasorBackbone=Ministral-3-14B-Reasoning-25122026.03 | 86.3 | 3,240 | 55.5 | |
| NoThinkingBackbone=Qwen3-14B2026.03 | 84.1 | 786 | 17.5 | |
| NoThinkingBackbone=Ministral-3-8B-Reasoning-25122026.03 | 83.2 | 1,908 | 28.1 | |
| LCPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.08 | 83.2 | 2,397 | — | |
| OriginalBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.08 | 83 | 5,665 | — | |
| TrEffBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.08 | 82.8 | 2,813 | — | |
| L1-MaxBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.08 | 82.2 | 3,374 | — | |
| CoDBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.08 | 81.8 | 4,788 | — | |
| L1-ExactBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.08 | 81.8 | 3,335 | — | |
| DEERBackbone=Qwen3-14B2026.03 | 80.9 | 2,501 | 56.2 | |
| NoThinkingBackbone=Qwen3-8B2026.03 | 80.7 | 809 | 16.1 | |
| DEERBackbone=Qwen3-8B2026.03 | 79.9 | 2,602 | 52 | |
| DynasorBackbone=Qwen3-14B2026.03 | 79.6 | 1,702 | 42.4 | |
| NoThinkingBackbone=Ministral-3-14B-Reasoning-25122026.03 | 79.1 | 535 | 11.7 | |
| DynasorBackbone=Qwen3-8B2026.03 | 78.3 | 1,850 | 41 | |
| DEERBackbone=Ministral-3-8B-Reasoning-25122026.03 | 71 | 3,791 | 60.3 | |
| DEERBackbone=Ministral-3-14B-Reasoning-25122026.03 | 69.8 | 4,279 | 74.6 |