Mathematical Reasoning on AIME 24 (Mean@16, Pass@16)
18.33Mean@16ADS
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ADSBackbone=Olmo3-7B-SFT2026.06 | 18.33 | 54.45 | |
| DOTS+RRBackbone=Olmo3-7B-SFT2026.06 | 15.28 | 50 | |
| DOTS+RRBackbone=Qwen3-4B-Base2026.06 | 14.09 | 34.45 | |
| ADSBackbone=Qwen3-4B-Base2026.06 | 12.5 | 33.33 | |
| GRPOBackbone=Qwen3-4B-Base2026.06 | 12.22 | 31.11 | |
| ADSBackbone=Qwen2.5-Math-1.5B2026.06 | 9.44 | 32.22 | |
| DOTS+RRBackbone=Qwen2.5-Math-1.5B2026.06 | 8.75 | 33.33 | |
| GRPOBackbone=Olmo3-7B-SFT2026.06 | 6.67 | 33.33 | |
| BaseBackbone=Qwen3-4B-Base2026.06 | 5.83 | 27.78 | |
| GRPOBackbone=Qwen2.5-Math-1.5B2026.06 | 5.63 | 34.44 | |
| BaseBackbone=Olmo3-7B-SFT2026.06 | 4.31 | 24.44 | |
| BaseBackbone=Qwen2.5-Math-1.5B2026.06 | 2.64 | 24.45 |