Hi 👋

I’m Adarsh Jamadandi, a CNRS doctoral researcher at IRISA, Rennes, working on diffusion models for graphs with Dr. Nicolas Keriven. I study the learning dynamics of discrete diffusion models for graph generation — when they generalize versus memorize — with the goal of building more efficient models that scale to large graphs.

Previously I was a research assistant at SprintML with Franziska Boenisch and Adam Dziedzic, where we built the first framework for studying memorization in GNNs.

I did my Master’s at Saarland University, with a thesis at the Relational ML Lab under Rebekka Burkholz on mitigating over-squashing and over-smoothing to improve GNN generalization.

I obtained my Bachelor’s in Electronics and Communication Engineering from India, with a thesis on video anomaly detection advised by Dr. Uma Mudenagudi.

My CV can be found here

Updates