| models: | |
| - model: nvidia/AceMath-7B-Instruct | |
| parameters: | |
| weight: 0.65 | |
| - model: deepseek-ai/DeepSeek-R1-Distill-Qwen-7B | |
| parameters: | |
| weight: 0.35 | |
| merge_method: task_arithmetic | |
| base_model: Qwen/Qwen2.5-Math-7B-Instruct | |
| parameters: | |
| lambda: 0.5 | |
| dtype: float16 | |