Eva Yi Xie | Princeton Neuroscience Institute

About Me

I am a 3rd year PhD candidate in Neuroscience at the Princeton Neuroscience Institute, supported by a Centennial Fellowship. My research seeks to understand the neural basis of decision-making, spanning levels of analysis from circuit-specific mechanisms to brain-wide characterization. I approach these questions primarily from a computational perspective, with theoretical insights directly tested and refined through cellular-level experiments. These days, I dive into designing new animal behaviors and collecting electrophysiological data myself to better connect with experiments as a theorist. I am primarily advised by Carlos Brody, and co-advised by Nathaniel Daw and Ben Eysenbach. I am a visiting scientist in Stefan Mihalas’ Theory & Modeling group within the Allen Institute.

I graduated from MIT in May 2024 as a double major S.B. in Computer Science (AI + Decision Making) and Mathematics (courses 6-4 and 18), with a concentration in Economics. During my undergrad, I worked on grid and place cell modeling with Ila Fiete. Before then, I worked with Tomaso Poggio and Earl Miller on studying interhemispheric transfer of working memory. Aside, I have explored ML research in Madry Lab at MIT CSAIL, and did R&D work at Meta, Microsoft AI, and IBM Research as an intern.

🧠💭 I co-organize the Annual Conference on Cognitive Computational Neuroscience (CCN) 2027 in Edinburgh, UK; see important dates here. I recently co-organized a NeurIPS 2025 workshop, Data on the Brain & Mind; contributions and other details of the workshop are avaliable here. A recording of the workshop is here.

I’m happy to chat if our research interests overlap. Additionally, I’ve been fortunate to benefit from many learning opportunities and mentorship along my path of science: if you’re a curious student seeking guidance, I’d be glad to connect. My only ask is that you pay it forward and support others when you’re in a similar position to do so.

Name & Pronunciation

I published under my official name Yi Xie and more recently Eva Yi Xie, but I commonly go by Eva. I often display both names together as Eva Yi Xie (pronounced: Yi-va Yi Shieh).

Research

Broadly, I view cognition as a computational process grounded in the basis of neurons and synapses, and I believe we can gain insights into these processes through models of neural networks. With the very recent development of large-scale neurophysiological recordings and connectome datasets, we now have unprecedented opportunities to investigate: How do multiple brain regions interact to support decision-making? To that end, my computational work spans the following three aspects, in complement to my ongoing development of new behavioral tasks in rats for studying how the brain detects and incorporates statistical structures into decision-making.

Normative: Developing biologically plausible models of interacting brain regions that perform brain-like computations, to uncover principles of neural coding and information flow that can be directly verified with neurophysiology experiments [Yi Xie, Jaedong Hwang, Carlos Brody, David Tank, Ila Fiete; ICML 2025]. The predictions on grid cells from this work are currently being experimentally validated, which we will present preliminary results at SfN 2026.
Mechanistic: Analyzing networks with biologically plausible properties, such as theoretically underexplored heavy-tailed synaptic weight distributions [Eva Yi Xie, Stefan Mihalas, Łukasz Kuśmierz; NeurIPS 2025] to reveal the dynamics and info-processing constraints imposed by connectivity structure alone. We are actively extending this work to make predictions and validations, utilizing connectomic and whole-brain imaging data in fruit flies.
Descriptive: Creating computational tools tailored for large-scale brain-wide recordings to decode multi-region coordinated neural signals and capture the moment-to-moment computations animals perform when making decisions (e.g., When precisely do animals make up their mind?) [Ongoing].

Together, these efforts aim to advance our understanding of multi-region neural computation and may also offer useful insights for developing AI systems that reflect key properties of the brain—such as robustness and efficiency.

CV

A complete CV is available upon request.