Kaiya Ivy Zhao

Kaiya Ivy Zhao

EECS PhD Student @ MIT

Massachusetts Institute of Technology

About

Hi! I’m Ivy, a PhD student at MIT CSAIL, where I am fortunate to be co-advised by Professor Phillip Isola and Professor Josh Tenenbaum. Previously, I worked with Professor Guangyu Robert Yang at MIT on autonomous agents for real-time interaction.

Since May 2026, I have also been a research intern at Adobe’s VideoAI Lab, working with Fabian Caba and Bryan Russell on video editing agents, representation learning, and personalization.

My research broadly lies at the intersection of representation learning, agentic systems, and human-centered AI.

I study how agents form and adapt representations of worlds, people, and other agents through interaction, and how those representations shape downstream behavior and generalization.

My current interests include:

  • Agents & Interactive Learning: How can agents learn from interaction, adapt to unfamiliar environments, and improve through experience and feedback?
  • World Models & Representation Learning: How can machines learn structured and generalizable models of the world? What representations support prediction, abstraction, and counterfactual reasoning?
  • Cognitive Science & Human-Centered AI: How does the human mind learn, reason, and adapt? How can these principles help AI systems better understand and adapt to individual people?

I often use games and creative workflows as concrete environments for studying these questions.

I publish under the name “Kaiya Ivy Zhao”, where “Zhao” is my surname and “Kaiya” is my forename. “Ivy” (ai·vee) is the name I usually go by.

Outside of research, I enjoy boba, musicals, classical music, snowboarding and fluffy cats.

Feel free to reach out via email to discuss research or collaborations.

Interests
  • Agents & Interactive Learning
  • World Models & Representation Learning
  • Cognitive Science & Human-Centered AI
Education
  • PhD in Computer Science, 2024 - present

    Massachusetts Institute of Technology

  • BSc in Computer Science, 2020 - 2024

    Fudan University

Selected Publications

For a complete and up-to-date list, please see my CV or Google Scholar.
(2026). Seeing Helps Reasoning in Language Models. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Findings, 2026.

Cite URL PDF

(2026). AI Gamestore: Scalable, Open-Ended Evaluation of Machine General Intelligence with Human Games. Conference on Neural Information Processing Systems (NeurIPS), Evaluations & Datasets Track, 2026.

Cite arXiv

(2026). HugAgent: Benchmarking LLMs for Simulation of Individualized Human Reasoning. Empirical Methods in Natural Language Processing (EMNLP), 2026 (Oral).

Cite arXiv

(2025). Assessing Adaptive World Models in Machines with Novel Games. Language, Agent, and World Models Workshop @ NeurIPS 2025 (Best Paper Award).

Cite arXiv

(2023). Lyfe Agents: Generative agents for low-cost real-time social interactions.

Cite arXiv