Resume-to-job matching is rarely evaluated independently in standard NLP benchmarks, and lacks a unified high-quality annotated dataset.