Mathematical Reasoning on AIME 2024 (Score, Best Score, ΔH)
0.47ScoreRLOO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RLOOBase Model=Qwen-3-8B, Max Context Length=4K2026.03 | 0.47 | 0.48 | -0.04 | |
| GRPOBase Model=Qwen-3-8B, Max Context Length=4K2026.03 | 0.45 | 0.47 | -0.07 | |
| REPO-RBase Model=Qwen-3-8B, Max Context Length=4K2026.03 | 0.45 | 0.46 | 0.13 | |
| GSPOBase Model=Qwen-3-8B, Max Context Length=4K2026.03 | 0.43 | 0.47 | 0.05 | |
| DAPOBase Model=Qwen-3-8B, Max Context Length=4K2026.03 | 0.43 | 0.44 | 0.23 | |
| ADAPOBase Model=Qwen-3-8B, Max Context Length=4K2026.03 | 0.43 | 0.43 | 0.13 |