Mathematical Reasoning on AMC (mean@8)
70.94Mean@8 ScorePrompt-GDRO
Evaluation Results
| Method | Links | |
|---|---|---|
| Prompt-GDROBackbone=Qwen3-8B-Base2026.01 | 70.94 | |
| Rollout-GDROBackbone=Qwen3-4B-Base2026.01 | 67.5 | |
| Rollout-GDROBackbone=Qwen3-8B-Base2026.01 | 66.88 | |
| GRPO (Baseline)Backbone=Qwen3-8B-Base2026.01 | 66.56 | |
| Prompt-GDROBackbone=Qwen3-4B-Base2026.01 | 64.06 | |
| GRPO (Baseline)Backbone=Qwen3-4B-Base2026.01 | 60.94 | |
| Prompt-GDROBackbone=Qwen3-1.7B-Base2026.01 | 39.38 | |
| Rollout-GDROBackbone=Qwen3-1.7B-Base2026.01 | 36.87 | |
| GRPO (Baseline)Backbone=Qwen3-1.7B-Base2026.01 | 34.69 |