Miles Wang

Wang studies how to oversee and evaluate increasingly capable models. His work spans chain-of-thought monitoring for reward hacking, frontier-risk and science evaluations, adversarial robustness, and agents that learn online for scientific research.