Retrieval & data
Reranking
Definition
Reranking takes an initial set of retrieved candidates and reorders them with a more accurate but slower model. It is one of the cheapest ways to materially improve RAG quality.
Retrieval is a two-stage problem. First, cheaply narrow millions of documents to perhaps 50 candidates using embedding similarity. Then apply a cross-encoder reranker, which examines the query and each candidate together and scores relevance far more accurately.
The reason for two stages is cost. Cross-encoders are too slow to run over an entire corpus but perfectly affordable over 50 candidates.
Teams debugging poor RAG answers often reach for a bigger generation model when the real problem is that the right passage was retrieved at rank 12 and never made it into the prompt. Reranking fixes that directly.
Related terms
Semantic search
Semantic search finds results by meaning rather than exact keywords, using embeddings to compare concepts. It matches "how do I get my money back" to a document titled "Refund Policy".
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.
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.
Knowledge cutoff
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.
Put this into practice
Understanding the term is step one. Our free courses and tools let you actually use it.