PhD student, Karlsruher Institut für Technologie
3 papers at NeurIPS 2025
We propose a meta-learning approach to enable Graph Network Simulators (GNSs) for fast adaptation to new physical parameters, leveraging conditional neural processes (CNPs) and neural operators to learn shared latent structures.
We use VLMs to generate high-level hand-object plans for dexterous manipulation.
We develop trust region methods for stochastic optimal control to improve sampling from unnormalized densities, transition path sampling, and diffusion model finetuning.
