2 ms·
Yes, I do admit that I simplified the mechanism in the article, but my question is why the scale of next-token prediction yields reasoning-like behavior.
by sichengo 8mo ago
Yes, I do admit that I simplified the mechanism in the article, but my question is why the scale of next-token prediction yields reasoning-like behavior.