👋 Hi, I’m Xinhong.

I am a fourth-year PhD student at the Institute for Brain and Cognitive Sciences and the Department of Automation at Tsinghua University. I am advised by Prof. Qionghai Dai and also work closely with Prof. Yuanlong Zhang.

I am an algorithm person at heart, driven by the long-term goal of understanding intelligence from both biological and artificial perspectives. To this end, I study machine learning, with a particular focus on its applications to neuroscience.

My current research interests currently include neural foundation models, reinforcement learning, and compression. My research experience spans neural data clustering, neural foundation models, microscopic data compression, and a decision-making project at the intersection of systems neuroscience and machine learning.

Research notes

Toward reusable models of neural population dynamics

I study how models of neural population activity can move beyond one recording session—reusing representations across experiments, datasets, tasks, and species.

Continuous autoregressive modeling for calcium traces

Calcium traces are continuous signals. CAPT asks whether modeling them directly can provide a simpler and more transferable interface than discretizing them into a vocabulary.

What should a neural foundation model transfer?

A foundation-style neural model should be judged not only by scale, but by what remains reusable: predictive structure, task information, and comparable representation spaces.

Selected publications

All publications →

arXiv preprint · 2026

CAPT: Continuous Autoregressive Population Transformer

A continuous autoregressive model for transferring calcium population dynamics across datasets, tasks, and species.

ICML 2026

CalM: A Self-Supervised Foundation Model for Calcium-imaging Population Dynamics

Self-supervised pretraining for neural forecasting, behavior decoding, and functional representation analysis.