Comment by Jirach05

5 months ago

Can anyone explain why these models decrease in performance on this "MCRC v2 (8-needle)" long context benchmark when thinking is turned on?

0 comments