(which has led me to believe that's a good approximation for hedonic adaptation, I've seen tons of attempts at demonstrating nerfing via benches, none persist)
"wow this model is really good I can push it so much further"
_pushes model further_
"Ugh why is this model failing now even though I'm asking it to do harder things and also got sloppier with my prompts because I got used to it being able to figure stuff out"
aabhay · · focus · HN ↗
refulgentis · · focus · HN ↗
(which has led me to believe that's a good approximation for hedonic adaptation, I've seen tons of attempts at demonstrating nerfing via benches, none persist)
vikramkr · · focus · HN ↗
_pushes model further_
"Ugh why is this model failing now even though I'm asking it to do harder things and also got sloppier with my prompts because I got used to it being able to figure stuff out"