Michael Bronstein

9 papers USI

• Amortized Sampling with Transferable Normalizing Flows
• Curly Flow Matching for Learning Non-gradient Field Dynamics
• Drag-and-Drop LLMs: Zero-Shot Prompt-to-Weights
• Equivariance Everywhere All At Once: A Recipe for Graph Foundation Models
• GradMetaNet: An Equivariant Architecture for Learning on Gradients
• Gradient Variance Reveals Failure Modes in Flow-Based Generative Models
• On Vanishing Gradients, Over-Smoothing, and Over-Squashing in GNNs: Bridging Recurrent and Graph Learning
• Over-squashing in Spatiotemporal Graph Neural Networks
• Progressive Inference-Time Annealing of Diffusion Models for Sampling from Boltzmann Densities