Mathematical Reasoning on MATH 500 (Acc, TR, s/q)
93.6AccuracyFull CoT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Full CoTBackbone=Llama-3.1-Nemotron-Nano-8B2026.05 | 93.6 | 0 | 1.07 | |
| PUMABackbone=Llama-3.1-Nemotron-Nano-8B2026.05 | 92.6 | 15.7 | 0.85 | |
| DEERBackbone=Llama-3.1-Nemotron-Nano-8B2026.05 | 91.8 | 22.7 | 1.83 | |
| DynasorBackbone=Llama-3.1-Nemotron-Nano-8B2026.05 | 91.4 | -9 | 0.99 | |
| DEERBackbone=DeepSeek-R1-Distill-Qwen-14B2026.05 | 91.2 | 26 | 3.68 | |
| PUMABackbone=DeepSeek-R1-Distill-Qwen-14B2026.05 | 91.2 | 21.3 | 1.84 | |
| Full CoTBackbone=DeepSeek-R1-Distill-Qwen-14B2026.05 | 91 | 0 | 2.09 | |
| DEERBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 90.6 | 39.6 | 1.63 | |
| Full CoTBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 90 | 0 | 0.98 | |
| PUMABackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 89.6 | 38.6 | 0.87 | |
| DynasorBackbone=DeepSeek-R1-Distill-Qwen-14B2026.05 | 89.2 | -2.3 | 8.59 | |
| DynasorBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 84.2 | 30.6 | 1.03 |