I’m currently a Machine Learning Researcher affiliated with the Princeton Robot Planning and Learning Group (PRPL) and the UBC Natural and Artificial Intelligence Lab (NAIL), working with Dr. Tom Silver and Dr. Kelsey Allen. I will soon begin my PhD under their supervision. Before this, I completed my Master’s in Computer Science at the University of Alberta under the supervision of Dr. Levi Lelis.

Currently, my research focuses on building agents that can learn physical strategies from only a few demonstrations. In particular, I study programmatic policies and learned abstractions that help agents generalize beyond the examples they observe. More broadly, I’m interested in how we can build systems that learn, reason, and understand the world more like humans, using ideas from symbolic reasoning, neurosymbolic AI, reinforcement learning, and imitation learning.

You can contact me at zb2882@princeton.edu

Publications

What to Represent, How to Act: Programmatic Feature Induction for Few-Shot Bayesian Imitation
_Zahra Bashir, Kelsey Allen, Tom Silver (Under Review for RLC - RL in Big Worlds Workshop)\

[Plastic Programming Languages: Learning Neuro-Augmented Domain-Specific Languages (In preparation)]
Zahra Bashir, David Aleixo, Kevin Ellis, Levi Lelis

SEGClobber - A Linear Clobber Solver
Taylor Folkerson, Zahra Bashir, Fatemeh Tavakoli, Martin Muller (International Computer Games Association)

Revisiting the Assessment of Programmatic Policy Interpretability: Insights from Human Evaluation
_Zahra Bashir, Michael Bowling, Levi Lelis _

LINT: Assessing the Interpretability of Programmatic Policies with Large Language Models
Zahra Bashir, Michael Bowling, Levi Lelis (RLC 2024 InterpPol Workshop)