v excited to finally announce our new work that formalizes one of the most effective practices for training LLMs—something that many industry leaders have conspicuously avoided discussing
19
97
895
Replies
s/o to
@danielking36
for the exceptional title. We also considered "Training on the test set is all you need", "The Unreasonable Effectiveness of Training on the Test Set", and "Intriguing Properties of Training on Test Data"
1
2
156
@leavittron
@jefrankle
Related
1
3
44
@leavittron
We have been working hard on this topic for serveral years
@znz8
@esruzzetti
. This is first version of our work
4
5
25
1
1
13
@leavittron
“Towards” means that you’re not there yet, so you’re still using the training set?
1
0
2
@leavittron
Thrilled to hear about this groundbreaking work! It's great to see a formalized approach to training LLMs, especially when it's been a less-discussed topic among industry leaders.
0
0
0
@leavittron
A database with NLP interface in sum.
Problem is: we don't know. They refuse to disclose what their training set is
0
0
2
