# @davidad — 2023-05-14

♥523 ↻82 · https://x.com/davidad/status/1657785914255368194

Biggest prosaic-LLM-alignment breakthrough of 2023 imo: turns out that, in GPT-2-XL, activation vectors in the residual steam have the same kind of affine structure as good old word2vec, but higher layers become emotional, then conceptual, then cognitivehttps://t.co/bzvUeGqFJG

tags: author:davidad, kind:tweet, on:gpt-2, year:2023
cited on: _dossiers/gpt-2.md, gpt-2