Mathematical Reasoning on MATH 500 (mean@8)
77.88Mean@8Rollout-GDRO
Evaluation Results
| Method | Links | |
|---|---|---|
| Rollout-GDROBackbone=Qwen3-8B-Base2026.01 | 77.88 | |
| Prompt-GDROBackbone=Qwen3-8B-Base2026.01 | 76.18 | |
| Prompt-GDROBackbone=Qwen3-4B-Base2026.01 | 75.78 | |
| Rollout-GDROBackbone=Qwen3-4B-Base2026.01 | 75.2 | |
| GRPO (Baseline)Backbone=Qwen3-8B-Base2026.01 | 73.45 | |
| GRPO (Baseline)Backbone=Qwen3-4B-Base2026.01 | 72.05 | |
| Rollout-GDROBackbone=Qwen3-1.7B-Base2026.01 | 63.98 | |
| Prompt-GDROBackbone=Qwen3-1.7B-Base2026.01 | 63.2 | |
| GRPO (Baseline)Backbone=Qwen3-1.7B-Base2026.01 | 50.62 |