Enthusiastically racing to remove the only practical way we have to monitor models intentions in exchange of a few tokenmaxxing percent increase. Can’t wait.
There is genuine research into latent space monitoring and interpretation - that's the only real way forward. CoT monitoring is a mirage and won't be useful for long, if it ever was
That's great but the sentiment that CoT isn't useful for monitoring doesn't seem to be shared by mech interp researchers like Nanda.
« If it ever was » is a wild claim considering our entire understanding of e.g. the HF attack comes from reading CoTs