1: \begin{abstract}

2: \jh{Meta-learning has showcased prowess in rapid adaptability, yet its reliance on fixed cardinalities contrasts starkly with human learning's inherent flexibility. This paper challenges the existing paradigm \nj{of meta-learning} by harnessing `label equivalence', emerging from stochastic numeric label assignments during episodic task sampling. We introduce the ``any-way" learning paradigm, an innovative model training approach free from fixed cardinality constraints. Surprisingly, this model not only matches but often outperforms traditional fixed-way models in terms of performance, convergence speed, and stability, disrupting long-held beliefs about domain generalization. To address the semantic information gap resulting from label equivalence, we further propose a mechanism to infuse semantic class information \nj{to the model}, enhancing the model's comprehension and functionality. Experiments on renowned architectures like MAML and ProtoNet affirm our method's effectiveness, pushing the boundaries of what defines ``true" meta-learning and setting the stage for a more versatile and semantically-aware future in the field.}

3: \end{abstract}

4: