A collection of Qwen2.5 models fine-tuned with Evolution Strategies using the Math dataset (levels 3-5).