2 papers
cs.LG2026
Improving scalable oversight with co-trained monitors
Joseph H. Rudoler, Kevin Tan, Benedict Tessler +2
Worker-monitor setups are a promising approach to AI oversight, but training workers against fixed monitors can incentivize monitor evasion. We study whether this failure mode can…
cs.CL2026
Model Hypnosis: Strong control of AI via additive subliminal effects
Enric Boix-Adsera, Benedict Tessler
We demonstrate that AI models are broadly susceptible to a phenomenon we call model hypnosis, in which individually weak and seemingly irrelevant cues in the prompt can be systemat…