Retrieve then rerank with Sentence Transformers
A local retrieval recipe that searches with a fast first stage and rescores candidate pairs with a cross-encoder.
Published by Sentence Transformers contributors
How to use it
Encode corpus passages once and fetch an initial set of candidates for each query. Send the candidate pairs through the cross-encoder and return the highest-scoring passages.
Log both rankings on a small judged query set. The first stage controls what can be recovered, while the reranker only changes the order of the candidates it receives.
- Compatibility
- Python, Sentence Transformers, PyTorch
- Components
- Two-stage retrieval, Cross-encoder reranking, Embedding search
- Access
- Local Python recipe. Download the selected model weights before offline runs.