Y
Hacker News
new
|
ask
|
show
|
jobs
by
hashta
261 days ago
I’m not sure AF3’s performance would hold up if it hadn’t been trained on data from AF2 which itself bakes in a lot of inductive bias like equivariance