The dataset is available at https://huggingface.co/datasets/weidawang/CMPhysBench. CMPhysBench is a benchmark for evaluating large language models in condensed matter ...