Zhimin Hu
School of Psychological and Brain Science & Interactive Computing
I am a first-year Ph.D. student in the Cognition and Brain Science Program at Georgia Tech, advised by Sashank Varma. I received my B.S. in Computer Science from the University of Wisconsin-Madison, where I worked with Gary Lupyan, Larissa Albantakis and Catherine Marler.
I identify as both a cognitive scientist and a computer scientist, curious about how biological and artificial minds compute. On the science side, I use behavioral experiments and computational modeling to study how humans reason and plan adaptively under resource constraints. On the engineering side, I tackle open problems in LLM reasoning and robotic learning. I am also especially drawn to using representational geometry to characterize how models and brains represent concepts across contexts and task demands.
Before joining Georgia Tech, I worked as a Research Scientist Intern at miHoYo developing the memory system for LLM agents. Before that, I was an Algorithm Engineer Intern at Z.ai working on the continual pretraining & SFT of GLM-4.
news
| Oct 01, 2026 | One paper is accepted to NeurIPS Interp4Discovery 2026 as 🏆Oral presentation. |
|---|---|
| Jun 11, 2026 | One paper is accepted to ICML Mech Interp 2026 as a 🏆Spotlight (23/801). |
| Apr 08, 2026 | Two papers are accepted to CogSci 2026. |
selected publications
- NeurIPS 2026Position: Mechanistic Interpretability Needs a Logic of InferenceNeurIPS - Interp4Discovery (🌟Oral), 2026
- ICML 2026Language Models Represent and Transform Concepts with Shared GeometryICML - Mech Interp (🌟Spotlight), 2026
- CogSci 2026The Representational Geometry of NumberIn Proceedings of the Annual Meeting of the Cognitive Science Society 48, 2026
- CogSci 2026Are More Tokens Rational? Inference-Time Scaling in Language Models as Adaptive Resource RationalityIn Proceedings of the Annual Meeting of the Cognitive Science Society 48, 2026
- Neural CompFrom Function to Implementation: Exploring Degeneracy in Evolved Artificial AgentsNeural Computation (🌟Featured on Cover), 2025
- Evolang XVFailures and Successes to Learn a Core Conceptual Distinction From the Statistics of LanguageIn The Evolution of Language: Proceedings of the 15th International Conference, 2024