Hacker News new | ask | show | jobs
by s314 33 days ago
You can actually de-censor an LLM without understanding how it works from a mechanistic perspective. (See R1 1776)

So I don't think there'll be effort to "obfuscate"