Geodesic Research
@GeodesResearch
We're behind
Building the base of alignment.
177 Following 831 Followers
In our new paper, we demonstrate that you can achieve selective generalisation of misalignment by midtraining on synthetic documents that describe how AIs can be misaligned in a special mode marked by a new token (a neologism), but remain aligned outside this mode.
Paper:
LessWrong post:
Show more
