Addison J. Wu

Currently, I'm...

• making sure AI development goes smoothly @ the Center for AI Safety, advised by Mantas Mazeika and Dan Hendrycks
• tinkering around @ Princeton Language and Intelligence, advised by Tom Griffiths and Zhuang Liu
  • teaching models to act fairly, honestly, cautiously, and strategically in the settings that matter
  • working out how data diversity and training constraints shape what models learn

Previously, I was...

• finding things by the sound they make @ the Multi-Physics Interaction Lab, University of Waterloo, advised by Jean-Pierre Hickey

I've co-led projects published at ICML (including a 2026 Oral), COLM, and NeurIPS
My work has also been featured in several press and TV outlets. Check out the rest at Google Scholar

I love to travel ✈️
🇨🇦 🇺🇸 🇲🇽 🇵🇷 🇵🇦 🇧🇷 🇯🇵 🇰🇷 🇨🇳 🇭🇰 🇮🇪 🇬🇧 🇫🇷 🇪🇸 🇧🇪 🇨🇭 🇮🇹 🇻🇦 🇸🇲 🇩🇰 🇸🇪 🇳🇴