Y
Hacker News
new
|
ask
|
show
|
jobs
by
rocho
569 days ago
That's correct. I saw a paper recently that showed how LLMs performance collapses when they are trained on synthetic data.