Explore our projects, publications and pilot experiments
Publishing rigorous empirical work for safe AI: evaluations, interpretability and more.
Research index.svg)
Our research is underpinned by novel approaches focused on neglected topics.
Our approach.svg)
Apart Sprints have kickstarted hundreds of pilot experiments in AI Safety.
Pilot experiments.png)
Our research focuses on critical research paradigms in AI Safety. We utilize interpretability to understand how AIs make decisions; process supervision techniques to ensure models behave as intended; benchmark tasks evaluate safety and alignment; and more. Apart produces foundational research enabling the safe and beneficial development of advanced AI.
