Mathematical Reasoning on AIME 2025 (reference set)
26.7pass@1ScaleBiO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ScaleBiOBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Fine-tuning=true, Training Samples=4K2024.06 | 26.7 | 36.7 | |
| UniformBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Fine-tuning=true, Training Samples=4K2024.06 | 20 | 33.3 | |
| LESSBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Fine-tuning=true, Training Samples=4K2024.06 | 20 | 36.7 | |
| RHO-LOSSBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Fine-tuning=true, Training Samples=4K2024.06 | 20 | 33.3 |