>release a great model
>cripple it post launch
>ship a new worse model
>claim its better

Follow

It becomes smarter when you say it's a benchmark :omegalul:

· · Web · 1 · 0 · 2

@matrix To be fair, these models are trained with the discussion on the faults of previous models. So chances are that if you ask a similar question just worded different, but using the same logical question and it might still shit the bed.

Sign in to participate in the conversation
Game Liberty Mastodon

Mainly gaming/nerd instance for people who value free speech. Everyone is welcome.