Mathematical Reasoning on Minerva Math (Pass@1)
38.2Pass@1SFT
Evaluation Results
| Method | Links | |
|---|---|---|
| SFTBase Model=Qwen2.5-Math-7B-Instruct, Evaluation Protocol=Zero-shot2025.10 | 38.2 | |
| LightRBase Model=Qwen2.5-Math-7B-Instruct, Evaluation Protocol=Zero-shot2025.10 | 34.2 | |
| BaselineBase Model=Qwen2.5-Math-7B-Instruct, Evaluation Protocol=Zero-shot2025.10 | 33.8 | |
| FREIABackbone=Qwen2.5-Math-1.5B-Instruct2026.04 | 32 | |
| LightRBase Model=Qwen2.5-Math-1.5B-Instruct, Evaluation Protocol=Zero-shot2025.10 | 32 | |
| SFTBase Model=Qwen2.5-Math-1.5B-Instruct, Evaluation Protocol=Zero-shot2025.10 | 31.6 | |
| GRPO (Supervised)Backbone=Qwen2.5-Math-1.5B-Instruct2026.04 | 31.3 | |
| FREIABackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 31.3 | |
| TTRLBackbone=Qwen2.5-Math-1.5B-Instruct2026.04 | 30.9 | |
| TTRLBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 30.9 | |
| GRPO (Supervised)Backbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 30.5 | |
| EntropyBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 29.8 | |
| EntropyBackbone=Qwen2.5-Math-1.5B-Instruct2026.04 | 29.4 | |
| IntuitorBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 29.4 | |
| BaselineBase Model=Qwen2.5-Math-1.5B-Instruct, Evaluation Protocol=Zero-shot2025.10 | 29.4 | |
| BaseBackbone=Qwen2.5-Math-1.5B-Instruct2026.04 | 28.7 | |
| IntuitorBackbone=Qwen2.5-Math-1.5B-Instruct2026.04 | 28.7 | |
| BaseBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 27.6 | |
| FREIABackbone=Qwen2.5-3B-Instruct2026.04 | 26.1 | |
| TTRLBackbone=Qwen2.5-3B-Instruct2026.04 | 25.8 | |
| GRPO (Supervised)Backbone=Qwen2.5-3B-Instruct2026.04 | 25.4 | |
| EntropyBackbone=Qwen2.5-3B-Instruct2026.04 | 25 | |
| BaseBackbone=Qwen2.5-3B-Instruct2026.04 | 24.3 | |
| IntuitorBackbone=Qwen2.5-3B-Instruct2026.04 | 24.3 | |
| SFTBase Model=DeepSeek-R1-Distill-Qwen-1.5B, Evaluation Protocol=Zero-shot2025.10 | 18.4 | |
| LightRBase Model=DeepSeek-R1-Distill-Qwen-1.5B, Evaluation Protocol=Zero-shot2025.10 | 18 | |
| BaselineBase Model=DeepSeek-R1-Distill-Qwen-1.5B, Evaluation Protocol=Zero-shot2025.10 | 16.2 | |
| SFTBase Model=Qwen2.5-Math-1.5B, Evaluation Protocol=Zero-shot2025.10 | 15.1 | |
| BaselineBase Model=Qwen2.5-Math-7B, Evaluation Protocol=Zero-shot2025.10 | 14 | |
| SFTBase Model=Qwen2.5-Math-7B, Evaluation Protocol=Zero-shot2025.10 | 12.1 | |
| LightRBase Model=Qwen2.5-Math-7B, Evaluation Protocol=Zero-shot2025.10 | 12.1 | |
| LightRBase Model=Qwen2.5-Math-1.5B, Evaluation Protocol=Zero-shot2025.10 | 11.4 | |
| BaselineBase Model=Qwen2.5-Math-1.5B, Evaluation Protocol=Zero-shot2025.10 | 9.9 |