Cath Ge-Wang

In this post, I will extend the conceptual and methodological work done previously on deployment-time misalignment continuation evals, and lay out a framework for us to think about misalignment spread. I'm currently an ERA fellow working with Alexandra Souly and Robert Kirk at UK AISI to build a multi-spread-vector misalignment...

Equal contributions: Cath Ge-Wang and Tyler Crosse Summary * AI control evaluations often model attackers as following a precomputed attack policy, such as an optimal attack insertion rate. However, a highly capable adversary may instead implement attack selection, strategically choosing to attack only when side-task success is highly likely. *...