Mathematical Reasoning on AMC 2023 (Mean@8)
95.3Mean@8DISA
Evaluation Results
| Method | Links | |
|---|---|---|
| DISABackbone=Qwen3-4B-Base2026.05 | 95.3 | |
| FlowRLBackbone=Qwen3-4B-Base2026.05 | 92.5 | |
| GSPOBackbone=Qwen3-4B-Base2026.05 | 83.5 | |
| GRPOBackbone=Qwen3-4B-Base2026.05 | 70 | |
| GRPOBackbone=Qwen2.5-7B2026.05 | 61.3 | |
| DISABackbone=Qwen2.5-7B2026.05 | 52.5 | |
| FlowRLBackbone=Qwen2.5-7B2026.05 | 50.6 | |
| GSPOBackbone=Qwen2.5-7B2026.05 | 48.1 | |
| VanillaBackbone=Qwen3-4B-Base2026.05 | 42.6 | |
| VanillaBackbone=Qwen2.5-7B2026.05 | 31.6 |