Mathematical Reasoning on AIME 2024 (Average@12 accuracy)
80.56Average@12 Accuracy (AIME 2024)ATESD
Evaluation Results
| Method | Links | |
|---|---|---|
| ATESDBackbone=Qwen3-8B2026.05 | 80.56 | |
| ATESDBackbone=Qwen3-4B2026.05 | 78.06 | |
| OPSDBackbone=Qwen3-8B2026.05 | 77.8 | |
| OPSDBackbone=Qwen3-4B2026.05 | 76.4 | |
| GRPOBackbone=Qwen3-8B2026.05 | 76.4 | |
| Base (Instruct)Backbone=Qwen3-8B2026.05 | 75.8 | |
| GRPOBackbone=Qwen3-4B2026.05 | 75.6 | |
| Base (Instruct)Backbone=Qwen3-4B2026.05 | 74.9 | |
| SFTBackbone=Qwen3-8B2026.05 | 72.3 | |
| SFTBackbone=Qwen3-4B2026.05 | 70.2 | |
| ATESDBackbone=Qwen3-1.7B2026.05 | 59.17 | |
| OPSDBackbone=Qwen3-1.7B2026.05 | 57.2 | |
| Base (Instruct)Backbone=Qwen3-1.7B2026.05 | 51.5 | |
| GRPOBackbone=Qwen3-1.7B2026.05 | 51.1 | |
| SFTBackbone=Qwen3-1.7B2026.05 | 48.4 |