YAN-LIU05/HCSU

HCSU: A Dataset and Benchmark for Fine-Grained Historical Calligraphy Style Understanding (V1.0) ECCV 2026 Yinsheng Yao*, Yan Liu*, Chen Ye† * Equal contribution. † Corresponding author. Abstract: Addressing the "Knowledgeable but unperceptive" dilemma where existing Large Vision-Language Models (LVLMs) possess historical knowledge but lack fine-grained calligraphy style perception, we introduce HCSU—the first large-scale dataset and evaluation benchmark specifically… See the full description on the dataset page: https://huggingface.co/datasets/YAN-LIU05/HCSU.