Cross-Model Similarity Mapping via Synthetic Query Probing

Synthetic Query Probing enables mapping similarity score distributions between different embedding models using linear, isotonic, or quantile mappings. This approach allows for better threshold portability and easier model migration in RAG systems without re-calculating all embeddings.