Mathematical Reasoning on AIME24 (pass@16)
71.7Pass@16DASD
Evaluation Results
| Method | Links | |
|---|---|---|
| DASDModel Scale=8B2026.05 | 71.7 | |
| HEPOModel Scale=8B2026.05 | 66.7 | |
| RLSDModel Scale=8B2026.05 | 62.3 | |
| GRPOModel Scale=8B2026.05 | 59.4 | |
| DASDModel Scale=4B2026.05 | 58.3 | |
| RLSDModel Scale=4B2026.05 | 56.5 | |
| GRPOModel Scale=4B2026.05 | 55 | |
| HEPOModel Scale=4B2026.05 | 53.3 | |
| SRPOModel Scale=8B2026.05 | 43.8 | |
| SRPOModel Scale=4B2026.05 | 40 | |
| GRPO(λ)-0.9Backbone=Qwen3-4B2026.05 | 38.83 | |
| RLSDModel Scale=1.7B2026.05 | 35.4 | |
| DASDModel Scale=1.7B2026.05 | 35 | |
| GRPOBackbone=Qwen3-4B2026.05 | 33.33 | |
| GRPOModel Scale=1.7B2026.05 | 32.5 | |
| HEPOModel Scale=1.7B2026.05 | 32.5 | |
| S-trace-0.9Backbone=Qwen3-4B2026.05 | 30 | |
| Qwen3Model Scale=8B2026.05 | 27.9 | |
| SDPOModel Scale=8B2026.05 | 27.9 | |
| OPSDModel Scale=8B2026.05 | 27.5 | |
| GRPO(λ)-0.99Backbone=Qwen3-4B2026.05 | 26.67 | |
| OPOBackbone=Qwen3-4B2026.05 | 25.48 | |
| SDPOModel Scale=4B2026.05 | 25 | |
| OPSDModel Scale=4B2026.05 | 24.2 | |
| P-trace-0.9Backbone=Qwen3-4B2026.05 | 23.33 | |
| SRPOModel Scale=1.7B2026.05 | 23.3 | |
| Qwen3Model Scale=4B2026.05 | 23.1 | |
| OPSDModel Scale=1.7B2026.05 | 15 | |
| SDPOModel Scale=1.7B2026.05 | 14.2 | |
| Qwen3Model Scale=1.7B2026.05 | 13.1 |