goatmeal [none/use name] to news • 1 year agoOver just a few months, ChatGPT went from accurately answering a simple math problem 98% of the time to just 2%, study findsexternal-linkmessage-square113 fedilinkarrow-up1112cross-posted to: chatgpt@lemmy.ml
arrow-up1112external-linkOver just a few months, ChatGPT went from accurately answering a simple math problem 98% of the time to just 2%, study findsgoatmeal [none/use name] to news • 1 year agomessage-square113 Commentsfedilinkcross-posted to: chatgpt@lemmy.ml
minus-squareShinji_Ikari [he/him]hexbear27·1 year agoI'm pretty sure the models' training data hasn't changed much yet, as the data set is huge. I'm not sure if this is negative tuning outcomes or what. link
I'm pretty sure the models' training data hasn't changed much yet, as the data set is huge.
I'm not sure if this is negative tuning outcomes or what.