Skip to main content
Technical implementation · AI Search Infrastructure

​
Definition

Zero-shot learning is a machine learning paradigm in which a model performs tasks it was not explicitly trained on — relying on generalized knowledge from pre-training to handle novel categories or tasks. LLMs exhibit zero-shot learning when they respond accurately to query types they have not seen explicit examples of. Zero-shot learning is why strong training corpus presence and clear entity signals matter beyond the specific queries a brand has been optimized for. An LLM with accurate, well-grounded knowledge of a brand can generalize that knowledge to answer novel query types — including queries no one anticipated. Brands with comprehensive entity records and broad content coverage benefit from zero-shot generalization to adjacent query types in ways that brands with narrow, specific optimization do not.

Pre-training

Foundation model

LLM brand memory

Training corpus

Semantic search

​
Relevant Plate Lunch Collective Services

AI SEO