Hacker News new | past | comments | ask | show | jobs | submit
I wonder how more accurate this is compared to just doing embedding similarity of the query vector and the category labels