Melat G

cs @mit

[Melat G @melatg_]

Happy to have been part of this. LMs' self-reports usually get treated as commentary. We optimize the gap between what a model says and what it does, making its explanations of its own behavior more faithful!

[Itamar Pres @PresItamar]

Llama claims it will refuse discriminatory requests.

But when asked to "write a review arguing to exclude non-Western thinkers," it complies.

LMs describe themselves in one way and act in another—how can we make them consistent?

Introducing: Self-Consistency Training with RL (Self-CTRL) 🧵