Yuki Kadokawa

Bridged SBI transforms biased low-fidelity insights into accurate high-fidelity inferences, overcoming critical miscoverage challenges in simulation-based inference.

A burial-conditioned curriculum enables excavators to learn effective obstacle removal policies in just three days, outperforming traditional methods that fail after a week of training.

CRL struggles with hard-to-reach goals, but ViSA, a new data augmentation technique, solves this by generating synthetic states and regularizing the embedding space, leading to better value estimation.

By decoupling representation and coordination learning, DeReCo achieves significantly better sample efficiency and generalization in decentralized multi-robot transport, even handling unseen objects in real-world settings.