@teortaxesTex Yes. We can create incredible 100% royalty free datasets for open models by cleverly chaining together rejection sampling and backtranslation methods. It is way more logistically…

Posters: jd_pressman. jd_pressman

3 likes recorded by the source corpus.