Zoe W. He

Zoe W. He

Hi! I am a Cognitive Science PhD candidate at UC San Diego, advised by Prof. Meenakshi Khosla. I study how AI systems represent and understand meaning across vision and language.

My research examines what semantic structure different models share, how their internal representations relate to their behavior, and how we can evaluate fine-grained understanding. I approach these questions through representational analysis, computational method development, and controlled experiments informed by cognitive science. More broadly, I am interested in developing a scientific understanding of intelligence, whether natural or artificial.

Previously, I worked with Prof. Angela J. Yu on human value-based decision-making under uncertainty, combining Bayesian computational modeling and pupillometry to understand how people adapt their decisions in changing environments.

News

04/2026 Gave a research talk on representational alignment between vision and language models at Cognition at the Shore, UC San Diego.
2026 Stability Implies Redundancy is accepted at ACL 2026!
02/2026 New preprint: Barycentric Alignment for Instance-Level Comparison of Neural Representations.
11/2025 Gave an invited talk at the Visual Learning Lab, UC San Diego.
11/2025 Presented Seeing Through Words, Speaking Through Pixels as an oral at EMNLP 2025.
08/2025 Presented a poster, Many-to-Many, Yet Convergent, at CCN 2025.

Selected Publications

Hover over a figure to see it larger, or click it to open the full image.

Probing Fine-Grained Semantic Alignment Across Vision and Language with Minimal Pairs
Zoe W. He, Yanyu Lei, Meenakshi Khosla
Under review
Barycentric Alignment for Instance-Level Comparison of Neural Representations
Shreya Saha*, Zoe W. He*, Meenakshi Khosla
Under review (* equal contribution)
Seeing Through Words, Speaking Through Pixels: Deep Representational Alignment Between Vision and Language Models
Zoe W. He, Sean Trott, Meenakshi Khosla
EMNLP 2025 (Oral)
Stability Implies Redundancy: Delta Attention Selective Halting for Efficient Long-Context Prefilling
Yujie Chen, Tailai Chen, Yifeng Gao, Zoe W. He, Yijue Xu, Shaobo Wang, Linfeng Zhang
ACL 2026
Pupil Size Reflects the Relevance of Reward Prediction Error and Estimation Uncertainty in Upcoming Choice
Zoe W. He, Maëva L’Hôtellier, Dalin Guo, Alexander Paunov, Florent Meyniel, Angela J. Yu
CogSci 2024 (Oral)
Multiple and Subject-Specific Roles of Uncertainty in Reward-Guided Decision-Making
Alexander Paunov, Maëva L’Hôtellier, Florent Meyniel, Dalin Guo, Zoe W. He, Angela J. Yu
eLife Reviewed Preprint (2024)

Other Publications & Presentations