Uncertainty in the Verbalizable Workspace
by Virginia Ceccatelli, Xiaoyu Liang, and Tobias Braun
Epistemic status: We found useful uncertainty signal in three reasoning models. All experiments were run on three seeds and are fully reproducible. Whether the same approach can become a broader tool for monitoring models is still an open question. Lingo might include “intend” or “conscious” and has to be taken...
Oct 177