Comment by seunosewa 10 days ago The degradation may be more significant within the day than at the same time every day. 1 comment seunosewa Reply GoatInGrey 10 days ago Sure, but it's still useful insight to see how it performs over time. Of course, cynically, Anthropic could game the benchmark by routing this benchmark's specific prompts to an unadulterated instance of the model.
GoatInGrey 10 days ago Sure, but it's still useful insight to see how it performs over time. Of course, cynically, Anthropic could game the benchmark by routing this benchmark's specific prompts to an unadulterated instance of the model.
Sure, but it's still useful insight to see how it performs over time. Of course, cynically, Anthropic could game the benchmark by routing this benchmark's specific prompts to an unadulterated instance of the model.