What is hybrid search and why is it better than pure vector search?
LLM & AILevel: IntermediateRetrieval Techniques
What is hybrid search and why is it better than pure vector search?
Hints
- Identify the stage: pretraining, adaptation, generation or evaluation.
- Name the inputs, constraints and a case where the approach fails.
Answer
Pure vector search excels at semantic similarity but fails on:
- Exact strings, codes, product names
- Rare domain-specific terms not in the embedding model's training data
- Abbreviations and acronyms
Pure keyword search (BM25) excels at exact matches but misses:
- Synonyms, paraphrases
- Semantic meaning across vocabulary
Hybrid search combines both:
BM25 results + Vector similarity results → Reciprocal Rank Fusion (RRF) → Merged ranking
RRF: score = Σ 1/(k + rank_i) — penalizes low ranks, rewards consistently high ranks across both methods.
Virtually all production RAG systems use hybrid search for better recall across all query types.
Source and licence: JustInCache & contributors — Awesome AI Interviews · Apache-2.0