smileysmiley9990/SCORE-Bench-Copy

Dataset Description SCORE-Bench is a curated collection of 224 diverse, real-world documents manually annotated by experts. It is designed to benchmark document parsing systems against true production-grade challenges. Unlike traditional academic datasets often composed of clean, digital-native PDFs, this benchmark specifically targets the complexity found in actual enterprise workflows. Note on replication: This dataset is a standalone benchmark released after the publication… See the full description on the dataset page: https://huggingface.co/datasets/smileysmiley9990/SCORE-Bench-Copy.