X · @teortaxesTex
· X / Twitter
The yet unconquered frontier for open weights LLMs:
The yet unconquered frontier for open weights LLMs:GDP: Kind of questions where GLM 5.2 trips over (so does Opus 4.6, only Opus 4.8 does a pretty good job):"Recall how in deep learning we update weights in proportion to loss gradient (-ve) with respective parameters. What this means is if gradient is less, weight chang