Anthony Baez
Welcome to my website! I am a graduate student at MIT in the MAS program, where I work in the Cyborg Psychology group under Prof. Pat Pataranutaporn in the Media Lab. I am also a recipient of the NSF Graduate Research Fellowship.
My research interests lie in improving our understanding and safety of LLMs using mechanistic interpretability. I am also interested in how we can apply mechanistic interpretability for better monitoring and control in human-AI interaction.
Outside of work, I enjoy traveling, learning other languages, hiking, videogames, and everything science fiction.
Email - Google Scholar - LinkedIn - Github
Selected Works
Dissociating the Internal Representations of Sycophancy in LLMs
Anthony Baez, Sheer Karny, and Pat Pataranutaporn
Mechanistic Interpretability Workshop @ ICML 2026
Multi-Turn Neural Transparency: Surfacing Neural Activations Improves User Calibration to LLM Behavioral Drift
Sheer Karny*, Anthony Baez*, and Pat Pataranutaporn
*Equal Contribution
Under Review
Guaranteeing Conservation of Integrals in Physics-Informed Neural Networks
Anthony Baez, Wang Zhang, Ziwen Ma, Subhro Das, Lam M Nguyen, and Luca Daniel
Arxiv Preprint