Strategy in training LLMs that optimizes the ratio of model size to training data size.
Generality: 275
