←back to thread

Macrodata Refinement

(lumon-industries.com)
722 points gaws | 1 comments | 01 Feb 25 21:46 UTC | HN request time: 0.392s | source
Show context
CharlesW ◴[01 Feb 25 22:41 UTC] No.42903210[source]▶
Please try to enjoy all comments equally, and not show preference for any over the others.
replies(6): >>42903450 #>>42904808 #>>42904831 #>>42907684 #>>42908326 #>>42911171 #
1. pizza ◴[02 Feb 25 10:14 UTC] No.42907684[source]▶
That’s the limiting state behavior of the global optimum GRPO trained language model, if you squint at it and look at it just right, funnily enough..