VDE Bench: Evaluating The Capability of Image Editing Models to Modify Visual Documents
VDE Bench is a rigorous benchmark designed to test how well image editing models can change dense text in visual documents, specifically focusing on bilingual Chinese and English texts. It involves creating 3,506 samples with human annotations for text and table edits. The evaluation checks if the edits are accurate regarding OCR output, structural fidelity, and visual preservation.