Abstract: The ability to generate unseen combinations of known components, known as compositional generalization, is crucial in machine translation. Although neural machine translation performs well ...
The agent acquires a vocabulary of neuro-symbolic concepts for objects, relations, and actions, represented through a ...
Abstract: Conventional federated learning (FL) assumes the homogeneity of models, necessitating clients to expose their model parameters to enhance the performance of the server model. However, this ...