Sparse Vectors and SPLADE (EXPLAINED)
Learned sparse representations, inverted index integration, lexical expansion, and fusion with dense ANN.
TL;DR — Quick Answer
Sparse vectors represent documents in high-dimensional lexical space with few non-zero weights — combining keyword specificity with learned term expansion. SPLADE learns sparse representations outperforming BM25 on many benchmarks. Store sparse vectors in engines supporting inverted indexes (Elasticsearch, Vespa, Milvus sparse) and fuse with dense ANN via RRF for hybrid retrieval.
The Interview Question
Explain sparse vectors and SPLADE in modern hybrid retrieval. How do they differ from dense embeddings?
Deep Explanation
Dense vs sparse
Sign in to unlock full answer
Get deep explanations, PDF export & all Vector Databases questions
- 21 more sections of deep explanation
- Real-world examples
- Common mistakes
- Interviewer expectations
- Follow-up questions
SPLADESparse VectorsHybrid SearchLexicalVespaElasticsearchPinecone