Is sampling at negative temperature¹ plausibly bad in terms of model welfare? The resulting output is the sequence of according-to-the model least likely tokens, and plausibly puts the model in a…

Posters: niplav_site. niplav_site

2 likes recorded by the source corpus.

Topics: AI & LLMs, AI safety

View Post: https://x.com/niplav_site/status/1893414124341633236