Model & agent builders
Model and agent builders need to understand how their AI behaves in cyber work.
Evaluations help developers improve models and agents, including those built on existing models. Training data helps developers build and improve their own models.