Tag
The paper proposes 'unit' as an explicit primitive in machine learning, where learning tasks declare persistent individuals, and supervised learning specializes to unit-conditioned response laws with tokenization.