Mathematical Reasoning on AMC 2023 (Accuracy, Avg)
86.09AccuracyCERO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CEROBackbone=Qwen3-4B-Instruct, Training Algorithm=CERO2026.06 | 86.09 | 50.11 | |
| GRPOBackbone=Qwen3-4B-Instruct, Training Algorithm=GRPO2026.06 | 85.46 | 46.83 | |
| CEROBackbone=Qwen2.5-Math-7B, Training Algorithm=CERO2026.06 | 67.03 | 30.66 | |
| GRPOBackbone=Qwen2.5-Math-7B, Training Algorithm=GRPO2026.06 | 65.31 | 29.13 | |
| CEROBackbone=R1-Distill-1.5B, Training Algorithm=CERO2026.06 | 63.43 | 26.84 | |
| CEROBackbone=Qwen3-4B-base, Training Algorithm=CERO2026.06 | 58.75 | 28.45 | |
| GRPOBackbone=R1-Distill-1.5B, Training Algorithm=GRPO2026.06 | 54.68 | 22 | |
| GRPOBackbone=Qwen3-4B-base, Training Algorithm=GRPO2026.06 | 50.46 | 22.41 |