Baidu's "Unlimited OCR" processes dozens of document pages in one pass by treating memory like human forgetting

Baidu has launched its Unlimited OCR technology, capable of reading dozens of document pages in a single pass, surpassing previous systems limited to ten pages. This innovation employs a modified attention mechanism that keeps memory usage flat, regardless of the number of pages processed. It currently ranks first on the most critical OCR performance benchmarks. This advancement significantly enhances document processing efficiency, paving the way for new applications in digital archiving and data analysis, thus improving how companies manage information.