I am a Ph.D. student in the Department of Electrical and Computer Engineering at Princeton University. I am very fortunate to be advised by Prof. Benjamin Eysenbach, and I am a member of Princeton RL lab.
I study how to build agents that can explore complex environments and learn useful skills without supervision, human guidance, or pre-collected datasets.
My recent work focuses on two core questions:
Unsupervised goal-reaching: How can an agent in a completely unknown environment discover its own intermediate skills and subgoals, then use them to reach difficult, faraway goals?
Learning world representations: In a big world full of overwhelming information, how should an agent decide what to pay attention to and what to discard? It is impossible for any agent to pay attention to everything due to computational and memory constraints, so agents should learn how to build useful abstractions of their environment.
Prior to this, my work centered on designing machine learning algorithms for decentralized financial markets. You can find a list of my publications from that line of research here. I received both my bachelor's and master's degrees in Electrical Engineering from Sharif University of Technology in Iran.
(ICML26) Learning to Perceive the World Through Control: Empowerment-Based Representation Learning. Mahsa Bastankhah, Sophie Broderick, Benjamin Eysenbach. Website, Paper
(ICLR26-Blogpost) Learning to Maximize Rewards via Reaching Goals. Chongyi Zheng, Mahsa Bastankhah, Grace Liu, Benjamin Eysenbach. Link
(ICLR26) Demystifying the Mechanisms Behind Emergent Exploration in Goal-conditioned RL. Mahsa Bastankhah*, Grace Liu*, Dilip Arumugam, Thomas L. Griffiths, and Benjamin Eysenbach, (*Equal contribution) Website, Paper