AeroJEPA: Learning Semantic Latent Representations for Scalable 3D Aerodynamic Field Modeling
Giral, …, Vinuesa et al. · arXiv 2026 · Latent models
What it does
Predicts the latent representation of the flow from geometry and conditions, and decodes a field only when one is needed.
Why it matters
Brings joint-embedding predictive training to aerodynamics, with a latent space that is meaningful on its own.