Zero-Shot Classification
Classifying content into categories the model was never explicitly trained on, by comparing the input's embedding to embeddings of category descriptions. For example, determining whether an email is about "billing," "technical support," or "sales" by comparing its embedding to those category labels.
Zero-Shot Performance
A model's ability to perform well on a task it was never specifically trained for. For embedding models, this means producing useful embeddings for new domains or tasks without additional fine-tuning. Strong zero-shot performance indicates broadly generalizable representations.