#vllm

PaDoc proposes parallel decoding for document parsing

The arXiv paper describes a layout-grounded MLLM parser that decodes document regions concurrently.