Mathematical Reasoning on AMC (avg@16)
75.1Avg@16Scaf-GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| Scaf-GRPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 75.1 | |
| Vanilla GRPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.10 | 65.4 | |
| Scaf-GRPOBackbone=Qwen2.5-Math-7B2025.10 | 63.8 | |
| Oat-ZeroBackbone=Qwen2.5-Math-7B2025.10 | 62.7 | |
| Vanilla GRPOBackbone=Qwen2.5-Math-7B2025.10 | 61.1 | |
| LUFFYBackbone=Qwen2.5-Math-7B2025.10 | 60.4 | |
| Scaf-GRPOBackbone=Qwen2.5-7B2025.10 | 54.2 | |
| Vanilla GRPOBackbone=Qwen2.5-7B2025.10 | 53 | |
| Scaf-GRPOBackbone=Qwen2.5-Math-1.5B2025.10 | 50.9 | |
| SimpleRL-ZeroBackbone=Qwen2.5-Math-7B2025.10 | 49.4 | |
| Vanilla GRPOBackbone=Qwen2.5-Math-1.5B2025.10 | 42.2 | |
| Scaf-GRPOBackbone=Llama-3.2-3B-Instruct2025.10 | 32.8 | |
| Vanilla GRPOBackbone=Llama-3.2-3B-Instruct2025.10 | 28.7 |