Mathematical Reasoning on MATH 500 (Acc, #Tok, LR)
95AccuracyNEAT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NEATModel=Qwen3-8B2026.02 | 95 | 3,906 | 24 | |
| VanillaModel=Qwen3-8B2026.02 | 94.6 | 5,140 | — | |
| NEATModel=Qwen3-14B2026.02 | 94.6 | 2,992 | 34.3 | |
| CGRSModel=Qwen3-14B2026.02 | 94.5 | 3,235 | 28.9 | |
| VanillaModel=Qwen3-14B2026.02 | 94.1 | 4,551 | — | |
| TALEModel=Qwen3-14B2026.02 | 93.7 | 3,389 | 25.5 | |
| CGRSModel=Qwen3-8B2026.02 | 93.3 | 3,507 | 31.8 | |
| TALEModel=Qwen3-8B2026.02 | 92.3 | 3,885 | 24.4 | |
| VanillaModel=DeepSeek-R1-Distill-Qwen-7B2026.02 | 92.2 | 3,743 | — | |
| NEATModel=DeepSeek-R1-Distill-Qwen-7B2026.02 | 92.2 | 2,914 | 22.1 | |
| CGRSModel=DeepSeek-R1-Distill-Qwen-7B2026.02 | 92 | 2,844 | 24 | |
| DynasorModel=Qwen3-8B2026.02 | 91.7 | 3,841 | 25.3 | |
| DEERModel=Qwen3-14B2026.02 | 91.7 | 1,956 | 57 | |
| DEERModel=DeepSeek-R1-Distill-Qwen-7B2026.02 | 89.6 | 2,272 | 39.3 | |
| TALEModel=DeepSeek-R1-Distill-Qwen-7B2026.02 | 89.1 | 2,657 | 29 | |
| DEERModel=Qwen3-8B2026.02 | 88.7 | 1,935 | 62.4 | |
| NoThinkingModel=Qwen3-8B2026.02 | 87.1 | 1,239 | 75.7 | |
| NoThinkingModel=Qwen3-14B2026.02 | 87 | 853 | 81.3 | |
| VanillaModel=DeepSeek-R1-Distill-Llama-8B2026.02 | 85.7 | 4,087 | — | |
| DynasorModel=DeepSeek-R1-Distill-Llama-8B2026.02 | 85 | 3,585 | 12.3 | |
| CGRSModel=DeepSeek-R1-Distill-Llama-8B2026.02 | 84.7 | 3,254 | 20.4 | |
| NEATModel=DeepSeek-R1-Distill-Llama-8B2026.02 | 84.7 | 2,934 | 28.2 | |
| DynasorModel=Qwen3-14B2026.02 | 84.4 | 3,667 | 19.4 | |
| TALEModel=DeepSeek-R1-Distill-Llama-8B2026.02 | 84 | 3,541 | 13.4 | |
| NoThinkingModel=DeepSeek-R1-Distill-Llama-8B2026.02 | 83.3 | 2,405 | 41.2 | |
| DEERModel=DeepSeek-R1-Distill-Llama-8B2026.02 | 82.3 | 2,722 | 33.4 | |
| DynasorModel=DeepSeek-R1-Distill-Qwen-7B2026.02 | 81.8 | 2,070 | 44.7 | |
| NoThinkingModel=DeepSeek-R1-Distill-Qwen-7B2026.02 | 80.9 | 1,173 | 65.4 |