Inverted File Index
Inverted File Index — A partitioned vector-search index that assigns vectors to coarse centroids and searches only selected partitions near the query.
Information-Retrieval Context
An IVF vector index learns coarse centroids and assigns vectors to those cells. At query time only selected nearby cells are scanned; parameters such as nprobe control the recall-latency trade-off, and product quantization can further reduce memory and distance-computation cost.
nlist controls the number of coarse partitions and nprobe controls the partitions visited per query. Increasing nprobe examines more candidates, affecting retrieval accuracy and search cost according to the data distribution. IndexIVFFlat and compressed IVF variants store candidates differently, so equal scan coverage does not imply equal memory use.
Terminology Boundary
This is not the same data structure as the classic inverted index used for term-based text search. The shared name reflects partitioned lookup, but IVF operates on vector-space cells rather than term posting lists.