Mathematical Reasoning on Minerva Math (Average@16 accuracy)
40.1Average Accuracy @16HiLL
Evaluation Results
| Method | Links | |
|---|---|---|
| HiLLBackbone=Qwen2.5-7B-Instruct2026.04 | 40.1 | |
| Scaf-GRPOBackbone=Qwen2.5-7B-Instruct2026.04 | 39.8 | |
| HiLLw/o TWBackbone=Qwen2.5-7B-Instruct2026.04 | 39.8 | |
| SAGEBackbone=Qwen2.5-7B-Instruct2026.04 | 39.3 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.04 | 39.1 | |
| LUFFYBackbone=Qwen2.5-7B-Instruct2026.04 | 39 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.04 | 38.1 | |
| HiLLBackbone=Llama-3.2-3B-Instruct2026.04 | 21.7 | |
| HiLLw/o TWBackbone=Llama-3.2-3B-Instruct2026.04 | 21.6 | |
| GRPOBackbone=Llama-3.2-3B-Instruct2026.04 | 20.5 | |
| SAGEBackbone=Llama-3.2-3B-Instruct2026.04 | 20.1 | |
| Scaf-GRPOBackbone=Llama-3.2-3B-Instruct2026.04 | 19.4 | |
| BaseBackbone=Llama-3.2-3B-Instruct2026.04 | 17.8 | |
| LUFFYBackbone=Llama-3.2-3B-Instruct2026.04 | 14.3 |