alt
Hacker News
vasco
•
today at 11:24 AM
•
0 replies
•
view on HN
Most data in the training set of most reasoning models is crap I guess.