Mathematical Reasoning on Minerva (Avg@32)
31.05Avg@32Dr.GRPO
Evaluation results
| Method | Links | |
|---|---|---|
| Dr.GRPOBackbone=Qwen3-4B2026.04 | 31.05 | |
| DSRLBackbone=Qwen3-4B2026.04 | 30.48 | |
| DSRLBackbone=Qwen3-8B2026.04 | 30.32 | |
| DAPOBackbone=Qwen3-4B2026.04 | 30.3 | |
| PPOBackbone=Qwen3-8B2026.04 | 30.3 | |
| RLOOBackbone=Qwen3-4B2026.04 | 30.27 | |
| PPOBackbone=Qwen3-4B2026.04 | 30.19 | |
| GRPOBackbone=Qwen3-4B2026.04 | 30.06 | |
| GRPOBackbone=Qwen3-8B2026.04 | 29.65 | |
| RLOOBackbone=Qwen3-8B2026.04 | 29.52 | |
| Dr.GRPOBackbone=Qwen3-8B2026.04 | 29.22 | |
| Reinforce++Backbone=Qwen3-8B2026.04 | 29.19 | |
| DAPOBackbone=Qwen3-8B2026.04 | 29.08 | |
| Reinforce++Backbone=Qwen3-4B2026.04 | 27.3 | |
| VanillaBackbone=Qwen3-4B2026.04 | 23.62 | |
| VanillaBackbone=Qwen3-8B2026.04 | 22.33 |