Hanlin Zhu

Emergence of Superposition: Unveiling the Training Dynamics of Chain of Continuous Thought

Auditing Black-Box LLM APIs with a Rank-Based Uniformity Test

GSM-Agent: Understanding Agentic Reasoning Using Controllable Environments

How Do Transformers Perform Two-Hop Reasoning in Context?

Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thought

Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers

Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning

Avoiding Catastrophe in Online Learning by Asking for Help

Watermarking using Semantic-aware Speculative Sampling: from Theory to Practice

Avoiding Catastrophe in Online Learning by Asking for Help