|
Some Recent Papers & Preprints
- Rational Transductors: New generative model augmenting self-attention Transformers with Weighted Automata.
- Self-Improvement: Principled LLM self-improvement based on coherence, with Jon Schneider and Yifan Wu.
- Modular LLM Learning: Theory and algorithms for modular learning of robust generative models, with Corinna Cortes and Yutao Zhong.
- Pion and Leon Optimization Algorithms: New principled optimization algorithms for nonsmooth nonconvex optimization, with Ruichen Jian, Zak Mhammedi, and Aryan Mokhtari.
- Swap regret and correlated equilibria beyond normal-form games. Best Paper Award EC 2025, with Eshwar Ram Arunachaleswaran, Natalie Collina, Yishay Mansour, Jon Schneider, and Balasubramanian Sivan.
- High-dimensional calibration from swap regret. NeurIPS 2025 Oral,
with Max Fishelson, Noah Golowich, and Jon Schneider.
Courses and Tutorials
|