Bio

I’m a final-year PhD student at the Computational and Biological Learning Lab (CBL) at Cambridge, supervised by Adrian Weller, and a research intern at Basis.

I’m on the job market for research roles in world models, RL, robotics and agents.

I received an MSc in Advanced Computer Science from Oxford and a BSc in AI from Edinburgh, where I graduated first in my cohort.

Interests

I am broadly interested in driving substantial advancements in humanity’s development. To this end, I pursue the complementary goals of understanding and developing generalist agents, drawing on diverse traditions in AI research, including formal languages, probabilistic modeling, machine learning, and planning.

Beyond this, I like to seek truth and beauty as they are found in nature, maths, and arts.

Selected Publications

arXiv 2026
EMPIRIC: Experiment-Driven Learning of Residual World Models for Robot Planning [Page] [Code]
Yichao Liang, Amber Li, Dat Nguyen*, Emily Bunnapradist*, Michelangelo Naim*, Sreela Kodali*, Matteo Merler, Bowen Li, Kiran Gopinathan, Yiyun Liu, Nikhil Pimpalkhare, Joshua B. Tenenbaum, Adrian Weller, Zenna Tavares, Tom Silver†, Kevin Ellis†
Extends a physics engine with learned code for the mechanisms it misses (e.g., glue curing, wind), infers their parameters from a few experiments, and plans with the result.
ICLR 2026(Top 1.9%)
ExoPredicator: Learning Abstract Models of Dynamic Worlds for Robot Planning [Blog] [Code]
Yichao Liang, Dat Nguyen, Cambridge Yang, Tianyang Li, Joshua B. Tenenbaum, Carl Edward Rasmussen, Adrian Weller, Zenna Tavares, Tom Silver*, Kevin Ellis*
Learns abstract world models of both the robot's actions and exogenous processes that unfold on their own (e.g., water heating, dominoes cascading) for long-horizon planning.
RSS 2026
KinDER: A Physical Reasoning Benchmark for Robot Learning and Planning [Page] [Code]
Yixuan Huang, Bowen Li, Vaibhav Saxena, Yichao Liang, Utkarsh Aashu Mishra, Liang Ji, Lihan Zha, Jimmy Wu, Nishanth Kumar, Sebastian Scherer, Danfei Xu, Tom Silver
Introduces 25 environments that isolate physical reasoning challenges (e.g., tool use, nonprehensile manipulation, dynamic constraints), where methods from TAMP to foundation models still struggle.
RA-L 2026
From Pixels to Predicates: Learning Symbolic World Models via Pretrained Vision-Language Models [Page] [Code]
Ashay Athalye*, Nishanth Kumar*, Tom Silver, Yichao Liang, Jiuguang Wang, Tomás Lozano-Pérez, Leslie Pack Kaelbling
Uses pretrained VLMs to propose and evaluate visual predicates, then learns a symbolic world model from a handful of demos that generalizes to novel goals and scenes.
NeurIPS 2025Spotlight
PoE-World: Compositional World Modeling with Products of Programmatic Experts [Page] [Code]
Wasu Top Piriyakulkij, Yichao Liang, Hao Tang, Adrian Weller, Marta Kryven, Kevin Ellis
Represents a world model as a product of many small LLM-synthesized programs, learning complex stochastic dynamics of Atari games from just a few observations.
ICLR 2025Spotlight
VisualPredicator: Learning Abstract World Models with Neuro-Symbolic Predicates for Robot Planning [Code]
Yichao Liang, Nishanth Kumar, Hao Tang, Adrian Weller, Joshua B. Tenenbaum, Tom Silver, João F. Henriques, Kevin Ellis
Invents neuro-symbolic predicates that combine symbolic structure with neural perception, learning abstract world models that plan sample-efficiently and generalize out of distribution.
CVPR 2024
Rapid Motor Adaptation for Robotic Manipulator Arms [Code]
Yichao Liang, Kevin Ellis, João F. Henriques
Brings Rapid Motor Adaptation to manipulation, inferring hidden object and environment properties (e.g., mass, friction, external forces) from recent history for robust, generalizable skills.

Recent Collaborators