Retrieval & data
Knowledge cutoff
Definition
A model's knowledge cutoff is the date after which it has no training data. It cannot know about events, releases or prices after that point unless given the information in the prompt or via search.
Training data is collected up to a point, so every model has a horizon. Ask about something after it and the model may say it does not know — or, less helpfully, produce a confident guess based on pre-cutoff patterns.
This is a common source of subtle errors: library versions, API signatures, pricing, and current officeholders all change. A model may describe a deprecated API perfectly because that API was current when it was trained.
Mitigations are retrieval, web search tools, or simply pasting current information into the prompt. For anything time-sensitive, verify against a live source.
Related terms
RAG (Retrieval-Augmented Generation)
RAG retrieves relevant passages from your own documents and inserts them into the prompt before the model answers. It grounds responses in your data, cuts hallucination, and needs no retraining.
Hallucination
A hallucination is model output that is fluent and confident but factually wrong — invented citations, non-existent functions, fabricated statistics. It stems from models being trained to produce plausible text, not verified truth.
Grounding
Grounding means tying model output to verifiable sources — retrieved documents, search results, or database records — rather than relying on parametric memory. Grounded answers can be checked.
Embedding
An embedding is a list of numbers representing the meaning of a piece of text, such that semantically similar texts have mathematically similar vectors. Embeddings make meaning-based search possible.
Vector database
A vector database stores embeddings and finds the most similar ones to a query vector quickly. It is the retrieval layer of most RAG systems. Examples include Pinecone, Weaviate, Qdrant and pgvector.
Chunking
Chunking splits documents into smaller passages before embedding them for retrieval. Chunk size is a key quality lever: too large dilutes relevance, too small loses the context needed to make sense.
Put this into practice
Understanding the term is step one. Our free courses and tools let you actually use it.