Mathematical Reasoning on MATH 500 (Accuracy @4, Average)
95.7Accuracy (avg@4)ASPO-Math-30B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ASPO-Math-30BBackbone=Qwen3-30B-A3B2025.10 | 95.7 | — | 98.4 | |
| DS-R1-Distill-Qwen-7B + IC-DAPOBackbone=Qwen-7B, Zero-shot mode=true, Time/Step (s)=315.62026.03 | 95.6 | 67.8 | — | |
| DS-R1-Distill-Qwen-7B + CE-GPPOBackbone=Qwen-7B, Zero-shot mode=true, Time/Step (s)=292.52026.03 | 95.3 | 67.3 | — | |
| GRPOBackbone=Qwen3-30B-A3B2025.10 | 95.3 | — | 97.8 | |
| DS-R1-Distill-Qwen-7B + DAPOBackbone=Qwen-7B, Zero-shot mode=true, Time/Step (s)=303.12026.03 | 94.1 | 65.3 | — | |
| DS-R1-Distill-Qwen-7B + GRPOBackbone=Qwen-7B, Zero-shot mode=true, Time/Step (s)=305.62026.03 | 93.7 | 61.4 | — | |
| DS-R1-Distill-Qwen-7BBackbone=Qwen-7B, Zero-shot mode=true2026.03 | 93.6 | 61.8 | — | |
| ASPO-Math-8BBackbone=Qwen3-8B2025.10 | 93.6 | — | 96.4 | |
| ASPO-Math-4BBackbone=Qwen3-4B2025.10 | 93.4 | — | 97 | |
| GRPOBackbone=Qwen3-8B2025.10 | 92.7 | — | 96 | |
| GRPOBackbone=Qwen3-4B2025.10 | 91.7 | — | 95.8 | |
| DS-R1-Distill-Qwen-1.5B + CE-GPPOBackbone=Qwen-1.5B, Zero-shot mode=true, Time/Step (s)=464.02026.03 | 91 | 55.7 | — | |
| DS-R1-Distill-Qwen-1.5B + IC-DAPOBackbone=Qwen-1.5B, Zero-shot mode=true, Time/Step (s)=477.22026.03 | 90.6 | 56.4 | — | |
| Nemotron-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 90.6 | — | 93.6 | |
| ASPO-Math-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 90.5 | — | 94.4 | |
| DS-R1-Distill-Qwen-1.5B + GSPOBackbone=Qwen-1.5B, Zero-shot mode=true, Time/Step (s)=437.32026.03 | 90.3 | 55.7 | — | |
| DS-R1-Distill-Qwen-1.5B + DAPOBackbone=Qwen-1.5B, Zero-shot mode=true, Time/Step (s)=459.62026.03 | 90 | 53.9 | — | |
| Qwen3-30B-A3BBackbone=Qwen3-30B-A3B2025.10 | 88.4 | — | 96 | |
| DS-R1-Distill-Qwen-1.5B + GRPOBackbone=Qwen-1.5B, Zero-shot mode=true, Time/Step (s)=457.42026.03 | 88.3 | 50.3 | — | |
| DeepScaleR-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 87.7 | — | 93.6 | |
| GRPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 87.6 | — | 94.6 | |
| DS-R1-Distill-Qwen-1.5BBackbone=Qwen-1.5B, Zero-shot mode=true2026.03 | 86 | 46.3 | — | |
| DS-R1-Distill-Qwen-1.5B + CISPOBackbone=Qwen-1.5B, Zero-shot mode=true, Time/Step (s)=466.32026.03 | 85.8 | 48.8 | — | |
| Qwen3-4BBackbone=Qwen3-4B2025.10 | 84.5 | — | 92.4 | |
| DeepSeek-R1-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 83.6 | — | 92.4 | |
| Qwen3-8BBackbone=Qwen3-8B2025.10 | 83.6 | — | 92.4 | |
| MCTS with Const-o-TBackbone=GPT-4.12025.10 | 83.1 | — | — | |
| MCTSBackbone=GPT-4.12025.10 | 82.2 | — | — | |
| CoVBackbone=GPT-4.12025.10 | 81.7 | — | — | |
| ToTBackbone=GPT-4.12025.10 | 80.8 | — | — | |
| MCTS with CoTBackbone=GPT-4.12025.10 | 80.2 | — | — | |
| CoTBackbone=GPT-4.12025.10 | 79.2 | — | — | |
| Const-o-TBackbone=GPT-4.12025.10 | 79.2 | — | — | |
| CoT + RSBackbone=GPT-4.12025.10 | 78.4 | — | — | |
| CTPBackbone=GPT-4.12025.10 | 78.3 | — | — | |
| MCTS with Const-o-TBackbone=LLaMA-32025.10 | 70 | — | — | |
| MCTS with CoTBackbone=LLaMA-32025.10 | 68.8 | — | — | |
| CoVBackbone=LLaMA-32025.10 | 68.2 | — | — | |
| CoT + RSBackbone=LLaMA-32025.10 | 66.6 | — | — | |
| MCTSBackbone=LLaMA-32025.10 | 66.2 | — | — | |
| CTPBackbone=LLaMA-32025.10 | 65 | — | — | |
| CoTBackbone=LLaMA-32025.10 | 64.2 | — | — | |
| Const-o-TBackbone=LLaMA-32025.10 | 64.2 | — | — | |
| ToTBackbone=LLaMA-32025.10 | 60.4 | — | — |