News for hackers. Stuff that matters.
Re:Transformer learning explained: Coinductive guide to indu... by adamnemecek on Tuesday February 28, 2023 @01:23PM (#34972842)
You are right, the paper can and will be improved. I think that I assume too much that people have read the Pang2014 paper.

Hopf coherence is not new, it's a name for an old concept that did not have a name.

Unit and counit are parts of the residual stream.

The antipode is a part of the OV circuit (6.2.2).

There are no comments on this comment yet.