AI & ML New Capability

A generative framework for graphs that closes the fidelity gap between energy-based models and discrete diffusion.

March 25, 2026

Original Paper

Graph Energy Matching: Transport-Aligned Energy-Based Modeling for Graph Generation

Michal Balcerak, Suprosana Shit, Chinmay Prabhakar, Sebastian Kaltenbach, Michael S. Albergo, Yilun Du, Bjoern Menze

arXiv · 2603.23398

The Takeaway

Energy-based models (EBMs) for discrete domains have historically been unstable and difficult to sample. GEM enables explicit relative likelihood modeling and property-constrained sampling for molecules, matching diffusion performance while providing better control for drug discovery.

From the abstract

Energy-based models for discrete domains, such as graphs, explicitly capture relative likelihoods, naturally enabling composable probabilistic inference tasks like conditional generation or enforcing constraints at test-time. However, discrete energy-based models typically struggle with efficient and high-quality sampling, as off-support regions often contain spurious local minima, trapping samplers and causing training instabilities. This has historically resulted in a fidelity gap relative to