RSS

LGOICOUR

Karma: 14

I am Neuroscience PhD turned software developer interested in alignment and mechanistic interpretability.

Toy trans­form­ers may rep­re­sent be­lief-state ge­om­e­try op­ti­mally but not minimally

LGOICOUR24 Jun 2026 4:08 UTC
16 points
0 comments3 min readLW link