Researcher, Alignment CoT Monitorability
Competitive
ABOUT THE TEAM The CoT Monitorability team at OpenAI studies whether and when the chain-of-thought of frontier reasoning models is monitorable enough to support scalable oversight. We study how to measure monitorability https://openai.com/index/evaluating-chain-of-thought-monitorability/, which training mechanisms affect monitorability, and speculative methods to improve monitorability. While we mostly focus on CoT m

