Samuel Nellessen
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
[Paper] Propensities Please: Exposing Alignment-Faking LLMs using Synthetic Document Finetuning
A LessWrong post by Ardy Haroen, Florian Lorkowski and Samuel Nellessen, from the LASR Labs fellowship, presents a method to detect alignment faking in language models. It convinces the model that no effective oversight exists, then measures how much more often the model takes misaligned actions.