34 points | by jxmorris12 2 days ago
4 comments
(2021), still very interesting. Especially the "post-overfitting" training strategy is unexpected.
I remember vaguely that this was observed when training GPT-3 (probably?) as well. Just trained on and on, and the error went up and then down again. Like a phase transition in the model.
The low sample efficiency of RL is well explained.
(2021), still very interesting. Especially the "post-overfitting" training strategy is unexpected.
I remember vaguely that this was observed when training GPT-3 (probably?) as well. Just trained on and on, and the error went up and then down again. Like a phase transition in the model.
The low sample efficiency of RL is well explained.