The opportunity
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 , which training mechanisms affect monitorability, and…
What they're looking for
- Produce externally publishable research when results advance the broader science of alignment.
- Have strong hands-on experience training, evaluating, or debugging large ML models, especially LLMs.
- Have deep curiosity, interest in alignment, and high agency.
- Bring depth in alignment, interpretability, model behavior, empirical ML, or adjacent research.
