I wonder what metrics can be used to detect the emergence of "model collapse" as described by Ross Anderson's group in this paper: https://www.lightbluetouchpa

I wonder what metrics can be used to detect the emergence of "model collapse" as described by Ross Anderson's group in this paper: https://www.lightbluetouchpaper.org/2023/06/06/will-gpt-models-choke-on-their-own-exhaust/ https://www.lightbluetouchpaper.org/2023/06/06/will-gpt-mode...