Machine Learning & Training

What Is Zero-Shot Learning?

Zero-shot learning is the ability of a model to perform tasks or recognize categories it has never seen during training. It relies on transferring knowledge, often through descriptions or shared representations, to handle new cases. Large language models frequently demonstrate zero-shot abilities on tasks they were not directly trained for.