Scientific Machine Learning

A non-autoregressive architecture combining DeepONets with DeepSets for in-context operator learning, achieving orders-of-magnitude parameter reduction and stronger noise robustness over transformer baselines.