Have been trying out RAG, and have been reading about Dense vs Sparse retrieval but my major question is how much of quality gets affected based on the selected embedding model I tried but with bunch of text the changes were minimal so I was not able to conclude it
Embedding model is important. But most of the issues i faced in my experience is due to prompt quality and business requirement mismatch. I would love to know more about your problem, you can dm me.
No problems just been playing a lot with RAG and wanted to know if embedding model has a bigger role in production grade data compared to all the different strategies, because most tutorials focus on strategies but none focuses on the embedding models even though there are quite a few
This is what a production pipeline would look like, yes embedding is important, but I never tried with different embeddings, to be honest, I was using OpenAI's Text-Embedding-3 family for embeddings.
Have been trying out RAG, and have been reading about Dense vs Sparse retrieval but my major question is how much of quality gets affected based on the selected embedding model I tried but with bunch of text the changes were minimal so I was not able to conclude it
Embedding model is important. But most of the issues i faced in my experience is due to prompt quality and business requirement mismatch. I would love to know more about your problem, you can dm me.
No problems just been playing a lot with RAG and wanted to know if embedding model has a bigger role in production grade data compared to all the different strategies, because most tutorials focus on strategies but none focuses on the embedding models even though there are quite a few
This is what a production pipeline would look like, yes embedding is important, but I never tried with different embeddings, to be honest, I was using OpenAI's Text-Embedding-3 family for embeddings.
Document parser/OCR
↓
Chunking / cleaning / metadata
↓
Embedding model
↓
Search/index layer
↓
Reranker
↓
LLM answer