Back at it—system gave us 500 gems… and 10× more junk 😂. Quick tweaks and we’re nearly done with stage one: mining pretrain data from rare, cross-domain PDFs.

#AIpretrain #SpanAware #TokenizerFree #PDFMining #XSpanformer #DataCuration #OpenScience
#artificalintelligence