Mathematical Reasoning on AIME 2024 (avg@64, pass@64)
61.7Average Score (avg@64)ASPO-Math-30B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ASPO-Math-30BBackbone=Qwen3-30B-A3B2025.10 | 61.7 | 90 | |
| GRPOBackbone=Qwen3-30B-A3B2025.10 | 59.5 | 80 | |
| ASPO-Math-8BBackbone=Qwen3-8B2025.10 | 52.4 | 83.3 | |
| ASPO-Math-4BBackbone=Qwen3-4B2025.10 | 50.5 | 83.3 | |
| GRPOBackbone=Qwen3-8B2025.10 | 50.3 | 83.3 | |
| ASPO-Math-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 49 | 80 | |
| Nemotron-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 48 | 76.7 | |
| GRPOBackbone=Qwen3-4B2025.10 | 43.4 | 83.3 | |
| GRPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 42.1 | 80 | |
| DeepScaleR-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 42 | 83.3 | |
| DeepSeek-R1-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 30.6 | 80 | |
| Qwen3-30B-A3BBackbone=Qwen3-30B-A3B2025.10 | 30.1 | 70 | |
| Qwen3-8BBackbone=Qwen3-8B2025.10 | 27 | 63.3 | |
| Qwen3-4BBackbone=Qwen3-4B2025.10 | 23.6 | 56.7 |