Training data consists of the examples (text, images, etc.) fed into a model during the learning process. The quality and quantity of this data fundamentally determine the model's capabilities.
The fuel for machine learning.
Data curation is now seen as just as important as algorithm design.