I bet it’s also like this in LLMs and we’re going to be seeing more diagrams that look like this from interpretability papers as the field matures

Posters: repligate. repligate

2 likes recorded by the source corpus.