rednote-hilab just released dots.ocr-1.5 on Hugging Face
A 3B-parameter multimodal OCR model that recognizes virtually any human script, achieves SOTA multilingual document parsing, and converts charts/diagrams directly into SVG code.
@firt How does it work?
Websites will have to fill in the content or will it be automatically created from the code?
I.e. old websites will have to adapt or they’ll be able to offer their content without code changes?
Chrome 146 includes an early preview of WebMCP, accessible via a flag, that lets AI agents query and execute services without browsing the web app like a user.
Services can be declared through an imperative navigator.modelContext API or declaratively through a form.
@akshay_pachaar https://t.co/olBdXmSMDa some folks are doing something similar in this paper.
Basically they have a curator agent that oversees the main agents and creates a 'cheatsheet' that can be added to the system prompt or can be used to curate the sys prompt itself!