Mathematical Reasoning on (MATH-500, Olympiad, Minerva, GSM8K, AMC, AIME24) (test)
58.12Overall ScorePretrainZero
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| PretrainZeroModel Series=Qwen3-30B-A3B-MoE-Base2025.12 | 58.12 | 81.7 | 43.4 | 41.2 | 94.4 | 70.62 | 17.4 | |
| PretrainZeroModel Series=Qwen3-8B-Base2025.12 | 57.72 | 81.9 | 42.5 | 43.4 | 93.5 | 65 | 20 | |
| Random RLPTModel Series=Qwen3-8B-Base2025.12 | 55.08 | 79.2 | 42.7 | 39 | 93.8 | 62.5 | 13.3 | |
| PretrainZeroRLPT steps=2000, RLVR steps=4002025.12 | 53.77 | 78.8 | 43 | 39.7 | 93 | 54.84 | 13.3 | |
| PretrainZeroModel Series=Qwen3-4B-Base2025.12 | 53.13 | 79.1 | 42.7 | 33.8 | 92.9 | 56.95 | 13.3 | |
| Random RLPTModel Series=Qwen3-30B-A3B-MoE-Base2025.12 | 52.62 | 79.2 | 41.2 | 38.6 | 82.4 | 59.77 | 14.58 | |
| Base ModelModel Series=Qwen3-30B-A3B-MoE-Base2025.12 | 52.49 | 74.7 | 43 | 22.8 | 91.1 | 66.95 | 16.36 | |
| PretrainZeroRLPT steps=1000, RLVR steps=4002025.12 | 51.84 | 77 | 43 | 32.4 | 92.5 | 55 | 11.13 | |
| Base ModelRLPT steps=0, RLVR steps=4002025.12 | 50.96 | 75.7 | 41.8 | 31.6 | 91.3 | 52.03 | 13.3 | |
| Random RLPTRLPT steps=2000, RLVR steps=4002025.12 | 47.87 | 74.8 | 38.5 | 22.1 | 87.5 | 54.3 | 10 | |
| Random RLPTModel Series=Qwen3-4B-Base2025.12 | 47.87 | 74.8 | 38.5 | 22.1 | 87.5 | 54.3 | 10 | |
| Base ModelModel Series=Qwen3-8B-Base2025.12 | 47.48 | 70.1 | 35.3 | 25.4 | 91.5 | 52.58 | 10 | |
| Base ModelModel Series=Qwen3-4B-Base2025.12 | 42.53 | 73.3 | 37.3 | 22.1 | 86.3 | 36.17 | 0 | |
| PretrainZeroModel Series=SmolLM3-3B-Base2025.12 | 38.03 | 62.6 | 25.6 | 22.1 | 83.7 | 27.5 | 6.7 | |
| Random RLPTModel Series=SmolLM3-3B-Base2025.12 | 35.95 | 59 | 21.5 | 20.2 | 82.5 | 32.5 | 0 | |
| Base ModelModel Series=SmolLM3-3B-Base2025.12 | 32.31 | 53.8 | 20.4 | 14 | 81.2 | 22.81 | 1.65 | |
| Continue PTModel Series=Qwen3-8B-Base2025.12 | 27.78 | 42.7 | 16.9 | 11.8 | 55.3 | 33.28 | 6.7 | |
| Continue PTModel Series=Qwen3-4B-Base2025.12 | 24.65 | 38 | 13.6 | 11 | 67 | 15 | 3.3 | |
| Supervised FTModel Series=Qwen3-8B-Base2025.12 | 19.23 | 30.5 | 11.7 | 15.4 | 32.8 | 25 | 0 | |
| Supervised FTModel Series=Qwen3-4B-Base2025.12 | 15.55 | 28.5 | 8.1 | 14.3 | 27.4 | 15 | 0 |