How do learned sparse retrievers such as SPLADE complement dense embeddings?
PICTURE THIS: AN LLM TURN
Simple meaning
SPLADE-style models expand queries and documents into sparse vocabulary-weighted vectors that still capture some synonyms.
WHY — Tokens instead of words?
LLMs use tokens (not full words) because it helps them:
question about Hybrid search.
trade-offs, and what you would actually do on a GenAI / LLM project - not buzzwords.
Each piece maps to a number the network can learn.
Fixed pieces are what transformers expect as input.
STEPS — What happens step by step?
Before you speak the answer, walk the interviewer through these steps:
- 1SPLADE-style models expand queries
and documents into sparse vocabulary-weighted vectors that still capture some synonyms.
- 2Combined with dense vectors
they cover exact terms and paraphrases in one sparse-plus-dense stack.
- 3You still need fusion
and a reranker for best nDCG.
- 4Give an example
One tiny concrete case you can say aloud.
- 5Common mistake
What juniors usually get wrong.
- 6Close
When you pick this over the alternative.
EXAMPLE — See it in action
Here's a short line you can speak, broken into clear beats:
Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.
Key takeaway
SPLADE-style models expand queries and documents into sparse vocabulary-weighted vectors that still capture some synonyms. Combined with dense vectors they cover exact terms and paraphrases in one sparse-plus-dense stack.