Interactive companions and demos built alongside my research.
An interactive companion to our ICML 2026 spotlight: a differential-equation GNN for dynamic graphs that learns exactly how much to remember and how much to forget. Play with the dimmer, build a colored-leaf-counting tree, and watch the lighthouse beam.
An interactive companion to our Graph Foundation Models Workshop paper at ICML 2026: a compositional framework that decouples time from space, injecting three frozen, domain-agnostic priors into a static GNN for transferable temporal graph learning. Watch the tide move over a static seabed, then explore the zero-shot transfer results.