Compositional Generalization for Real-World Robot Learning

A half-day workshop focusing on compositional or modular learning in robotics. The workshop will cover the application of modular systems to learning problems, exploring how the community can design benchmarks that evaluate compositional generalization across novel modalities, embodiments, and goal combinations. Targeting researchers in robotics, machine learning, and embodied AI, this workshop encourages innovative approaches that leverage scalable systems while retaining modular structure.