Mathematical Reasoning on Math500 (Accuracy, Avg, Speedup)
87.18AccuracySTRIDE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| STRIDEBase Model=Qwen3-4B-Base2026.06 | 87.18 | 49.24 | — | |
| HET-DAPOBase Model=Qwen3-4B-Base2026.06 | 86.96 | 47.27 | — | |
| DAPOBase Model=Qwen3-4B-Base2026.06 | 86.48 | 46.63 | — | |
| HICRA-GRPOBase Model=Qwen3-4B-Base2026.06 | 86.43 | 47.09 | — | |
| GRPOBase Model=Qwen3-4B-Base2026.06 | 83.75 | 45.28 | — | |
| GRPO + ARROLBackbone=Qwen-3-8B-Base, Method=GRPO + ARROL2026.03 | 81.45 | 59.53 | 1.62 | |
| GRPOBackbone=Qwen-3-8B-Base, Method=GRPO2026.03 | 81.25 | 56.66 | — | |
| STRIDEBase Model=Qwen2.5-7B-Instruct2026.06 | 80.17 | 44.71 | — | |
| GRPO + ARROLBackbone=Qwen-3-4B-Base, Method=GRPO + ARROL2026.03 | 80.04 | 53.54 | 1.63 | |
| GRPOBackbone=Qwen-3-4B-Base, Method=GRPO2026.03 | 79.64 | 51.01 | — | |
| HET-DAPOBase Model=Qwen2.5-7B-Instruct2026.06 | 78.55 | 42.18 | — | |
| DAPOBase Model=Qwen2.5-7B-Instruct2026.06 | 78.25 | 41.8 | — | |
| GRPOBase Model=Qwen2.5-7B-Instruct2026.06 | 78.07 | 41.61 | — | |
| HICRA-GRPOBase Model=Qwen2.5-7B-Instruct2026.06 | 77.58 | 42.01 | — | |
| BaseBase Model=Qwen3-4B-Base2026.06 | 63.8 | 29.4 | — | |
| BaseBase Model=Qwen2.5-7B-Instruct2026.06 | 62.4 | 34.1 | — | |
| GRPO + ARROLBackbone=Qwen-3-1.7B-Base, Method=GRPO + ARROL2026.03 | 62.3 | 37.09 | 1.61 | |
| GRPOBackbone=Qwen-3-1.7B-Base, Method=GRPO2026.03 | 60.89 | 34.79 | — | |
| STRIDEBase Model=Llama-3.1-8B-Instruct2026.06 | 57.19 | 25.87 | — | |
| HET-DAPOBase Model=Llama-3.1-8B-Instruct2026.06 | 55.48 | 23.97 | — | |
| HICRA-GRPOBase Model=Llama-3.1-8B-Instruct2026.06 | 55.25 | 23.3 | — | |
| DAPOBase Model=Llama-3.1-8B-Instruct2026.06 | 55.15 | 23.41 | — | |
| GRPOBase Model=Llama-3.1-8B-Instruct2026.06 | 54.22 | 22.92 | — | |
| BaseBase Model=Llama-3.1-8B-Instruct2026.06 | 50.2 | 17.78 | — | |
| GRPO + ARROLBackbone=Llama-3.2-1B-Instruct, Method=GRPO + ARROL2026.03 | 29.03 | 17.49 | 1.67 | |
| GRPOBackbone=Llama-3.2-1B-Instruct, Method=GRPO2026.03 | 24.2 | 14.63 | — |