Research
Notation: (α) authors listed alphabetically; when marked with †, only the co-first authors are listed alphabetically † equal contribution (r) randomized order among authors marked with †
What Is Lost in Post-Training? Default Collapse and the Loss of In-Context Steerability Across Diverse Perspectives
NeurIPS 2026 Workshop on Foundations of LLM Post-Training in Changing Environments
How RLHF Amplifies Sycophancy
AI Alignment From Social Choice Perspectives
July 2026
Robust AI Evaluation through Maximal Lotteries
Incentives in Federated Learning with Heterogeneous Agents
Generative Social Choice
Extended version of the paper published at EC 2024
Pairwise Calibrated Rewards for Pluralistic Alignment
SOAP: Improving and Stabilizing Shampoo using Adam
A New Perspective on Shampoo’s Preconditioner
Spotlight Top 2.1% of submissions
Axioms for AI Alignment from Human Feedback
Spotlight Top 2.1% of submissions
Learning Social Welfare Functions
Bias Detection Via Signaling
Optimal Engagement-Diversity Tradeoffs in Social Media
No publications match the current filters.