Mathematical Reasoning on AIME 2024 (Pass@64, Avg@64, Entropy)
34.38Average Score @64Clip-Free
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Clip-FreeClipping Strategy=Free2025.11 | 34.38 | 66.67 | 6.745 | |
| Pos-Adv-Reweight (Entropy-guided)Reweighting Strategy=Entropy-guided2025.11 | 34.38 | 73.33 | 18.746 | |
| Rand-Pos-Clip2025.11 | 34.27 | 66.67 | 5.763 | |
| DAPONupdate=42025.11 | 33.96 | 66.67 | 47.9 | |
| Ada-Ent-Regdelta (δ)=0.36572025.11 | 33.96 | 66.67 | 30.941 | |
| KL-Cov2025.11 | 33.96 | 66.67 | 19.695 | |
| Clip-HigherClipping Strategy=Higher2025.11 | 33.33 | 60 | 53.91 | |
| Ada-Ent-Regdelta (δ)=0.22025.11 | 32.92 | 66.67 | 19.692 | |
| Pos-Adv-Reweight (Epoch-wise)Reweighting Strategy=Epoch-wise2025.11 | 32.34 | 66.67 | 5.203 | |
| Clip-TighterClipping Strategy=Tighter2025.11 | 32.19 | 63.33 | 4.681 | |
| Clip-CovClipping Strategy=Covariance-based2025.11 | 31.98 | 70 | 20.899 | |
| Pos-Adv-Reweight (Stage-based)Reweighting Strategy=Stage-based2025.11 | 31.72 | 56.67 | 32.983 | |
| GRPONupdate=42025.11 | 31.41 | 63.33 | 1.789 | |
| Entropy-Adv2025.11 | 31.09 | 66.67 | 1.669 | |
| Adv <= 02025.11 | 29.79 | 63.33 | 88.384 | |
| GRPONupdate=22025.11 | 29.58 | 70 | 2.286 | |
| GRPONupdate=12025.11 | 28.75 | 63.33 | 11.838 | |
| Clip-LowerClipping Strategy=Lower2025.11 | 27.76 | 56.67 | 1.577 | |
| Adv >= 02025.11 | 27.55 | 53.33 | 1.46 | |
| Qwen2.5-Math-7B2025.11 | 10 | 60 | — |