Question

Your production RAG system uses a general-purpose embedding model with 1,536 dimensions, and its vector index in Azure AI Search contains millions of chunked documents. Evaluation shows poor retrieval relevance for your specialized medical terminology. You fine-tune a domain-specific embedding model that outputs 768 dimensions and it scores significantly higher on your relevance benchmark. What must you do before the fine-tuned embedding model can serve production retrieval queries?

Accepted Answer

Re-embed all existing documents with the fine-tuned model and rebuild the vector index with the new 768-dimension vector field, then embed incoming queries with the same model.

Answer

Update only the query-time embedding call to use the fine-tuned model, leaving the existing document vectors unchanged since the index already contains the content.

Answer

Pad the fine-tuned model's 768-dimension output to 1,536 dimensions with zeros so it remains compatible with the existing index schema.

Answer

Enable hybrid search so the keyword component compensates for the dimension mismatch between the query and document vectors.