Lune

Profiles2023-2026

Hanlin Zhu

University of California

10Papers
0Last author
26Unique co-authors
6First author
0Solo

First and last author come from byline position, which some venues assign alphabetically.

Recent papers10

  1. Auditing Black-Box LLM APIs with a Rank-Based Uniformity TestAuthor 4 of 9ICLR 2026 · 22 citations
  2. Emergence of Superposition: Unveiling the Training Dynamics of Chain of Continuous ThoughtAuthor 1 of 9ICLR 2026 · 21 citations
  3. Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous ThoughtAuthor 1 of 8NeurIPS 2025 · 86 citations
  4. Generalization or Hallucination? Understanding Out-of-Context Reasoning in TransformersAuthor 3 of 9NeurIPS 2025 · 10 citations
  5. Towards a Theoretical Understanding of the 'Reversal Curse' via Training DynamicsAuthor 1 of 9NeurIPS 2024 · 37 citations
  6. On Representation Complexity of Model-based and Model-free Reinforcement LearningAuthor 1 of 3ICLR 2024 · 6 citations
  7. Learning Personalized Alignment for Evaluating Open-ended Text GenerationAuthor 3 of 9EMNLP 2024 · 2 citations
  8. Importance Weighted Actor-Critic for Optimal Conservative Offline Reinforcement LearningAuthor 1 of 3NeurIPS 2023 · 21 citations
  9. Provably Efficient Offline Goal-Conditioned Reinforcement Learning with General Function Approximation and Single-Policy ConcentrabilityAuthor 1 of 3NeurIPS 2023 · 7 citations
  10. Optimal Conservative Offline RL with General Function Approximation via Augmented LagrangianAuthor 2 of 5ICLR 2023
Dusk over the sea between two cliffs drawn in fine vertical lines