Han-yu (Henry) Wang
Applied Artificial Intelligence undergraduate at the University of Hong Kong, working where cognitive science meets AI. My research runs along two lines: how people learn and remember, and what goes on inside language models when they reason, studied through behavioral experiments, model internals, and computational modeling.
Publications & preprints
Human learning & cognition
- The Resolving Power of Transfer Experiments: When Can Designs Distinguish Accounts of Generalization? (under review)
- Functional Encoding and Representational Binding in Componential Transfer (under review)
- When Does a Smooth Learning Curve Mean Smooth Learning? Representational Format, Identifiability, and the Bounded Averaging Artifact (under review)
- Humans Disengage, Reasoning Models Persist: Separating Difficulty Registration from Deliberation Allocation (under review)
Language models: mechanisms & behavior
- BRIGHT: A Realistic and Challenging Benchmark for Reasoning-Intensive Retrieval (ICLR 2025 Spotlight)
- Priors Persist Through Suppression: A Stroop Paradigm for Lexical Override (under review)
- Function-Vector Heads Are Two Populations: Writers and Cancellers in In-Context Learning (under review)
- When More Becomes Less: Position-Dependent Repetition Effects in Language Models (under review)
CV (PDF) · GitHub · LinkedIn · whenry6688@gmail.com
This site is interactive and needs JavaScript for the full experience.