Radar Telegram alerts Open Radar

Xiaomi introduced Xiaomi-OCR-0, a compact artificial intelligence model for working with documents that surpasses larger competitors by several times. The model has only 800 million parameters: it recognizes text in scans and photographs, analyzes tables and formulas, extracts data from invoices and receipts in JSON and answers questions about documents; In the OmniDocBench test, the model received 96.83 points - only 0.04 points less than the absolute leader; Among models with sizes of up to a billion parameters, it took first place. Everything works locally, without clouds:.

תרגם לעברית

Адзер К. Хроники Апокалипсиса. · · Source

Xiaomi introduced Xiaomi-OCR-0, a compact artificial intelligence model for working with documents that surpasses larger competitors by several times. The model has only 800 million parameters: it recognizes text in scans and photographs, analyzes tables and formulas, extracts data from invoices and receipts in JSON and answers questions about documents; In the OmniDocBench test, the model received 96.83 points - only 0.04 points less than the absolute leader; Among models with sizes of up to a billion parameters, it took first place. Everything works locally, without clouds:.

Xiaomi introduced Xiaomi-OCR-0, a compact artificial intelligence model for working with documents that surpasses larger competitors by several times. The model has only 800 million parameters: it recognizes text in scans and photographs, analyzes tables and formulas, extracts data from invoices and receipts in JSON and answers questions about documents; In the OmniDocBench test, the model received 96.83 points - only 0.04 points less than the absolute leader; Among models with sizes of up to a billion parameters, it took first place. Everything works locally, without clouds:.

More on Radar

Want to see every report in real time?

Hundreds of sources worldwide, translated into your language, with alerts for what matters to you.

Sign up free
Share on WhatsApp Get alerts on Telegram