Paper summary
Paper summary: SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion
SmolDocling is a 256M-parameter vision-language model that converts whole document pages end to end into DocTags, a markup capturing content, structure and location of every element.
- One tiny model replaces a pipeline of specialised models: layout, tables, code, equations and charts all come out as DocTags.
- It's reported to compete with VLMs up to 27 times larger.