Song Mei

4 papers University of California, Berkeley

• A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics
• Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers
• OVERT: A Benchmark for Over-Refusal Evaluation on Text-to-Image Models
• Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning