3 ms·
>the 70B variant can already surpass gpt-3.5-turbo-16k's overall performance on a suite of long-context tasks Now I believe Meta when they say this, as I can b
by brucethemoose2 3y ago
>the 70B variant can already surpass gpt-3.5-turbo-16k's overall performance on a suite of long-context tasks
Now I believe Meta when they say this, as I can believe they were careful about dataset contamination.