Y
Hacker News
new
|
ask
|
show
|
jobs
by
Ukv
438 days ago
Human win rate would be 1 minus the model win rate, to my understanding. So 77% against ELIZA, 27% against GPT-4.5 with a human persona.