Princeton SuperAlignment

Measuring Intelligence Beyond Human Scale

Jerry Han, Rafael Moschopoulos, Ella Colby, Vishrut Goyal, Andrew Tu, Kia Ghods, Mark Braverman, Elad Hazan, 2026.

A relative-evaluation framework where models generate public challenges for other systems, enabling scalable measurement of capabilities beyond human-authored benchmarks.

Project page · arXiv

AI Alignment via Incentives and Correction

Rohit Agarwal, Joshua Lin, Mark Braverman, Elad Hazan, 2026.

A mechanism-design view of alignment in solver-auditor AI pipelines, where reward design induces the behavioral equilibrium of both solving and oversight.

arXiv

Playing Large Games with Oracles and AI Debate

Xinyi Chen, Angelica Chen, Dean Foster, Elad Hazan, 2023.

Oracle-based algorithms for regret minimization in very large games, motivated by AI Safety via Debate and language-based action spaces.

arXiv

The Hidden Game Problem

Gon Buzaglo, Noah Golowich, Elad Hazan, 2025.

A study of efficient regret minimization in games with hidden high-reward structure, motivated by AI alignment and language games.

arXiv

The Statistical Orthogonality Thesis

Elad Hazan, Helen Qu, Lauren Li, Minimizing Regret, 2026.

A statistical perspective on orthogonality in AI systems and its implications for alignment.

Minimizing Regret

AI Alignment as Equilibrium Design

Elad Hazan, LessWrong, 2026.

An accessible overview of the mechanism-design perspective on AI alignment.

LessWrong