Estimating no-CoT task completion time horizons of LLMs

​Many efforts to ensure frontier AI models are safe rely on monitoring their chain-of-thought (CoT) reasoning. If models become able to perform sufficiently complex reasoning internally, without explicit thinking tokens, this would undermine such oversight.

​Co-authors Rauno Arike and Josh Hill will discuss "Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models", a recent paper which measured frontier models' capabilities to complete tasks without CoT reasoning.