VDE Bench: Evaluating The Capability of Image Editing Models to Modify Visual Documents

VDE Bench (Visual Doc Edit Bench) has been introduced as a benchmark for assessing image editing models' capabilities in modifying dense visual documents, particularly those containing bilingual Chinese-English text. It features a dataset of 942 annotated editing samples from various document types, including academic papers and newspapers, and employs a novel evaluation framework that quantifies editing performance at the OCR parsing level. This benchmark is significant for practitioners as it addresses the limitations of existing models in handling complex, densely populated text documents and non-Latin scripts, thereby paving the way for improved image editing applications in diverse linguistic contexts.