RAG and Similarity Metrics: Why not use more advanced methods?

Beginners
1

Hi everyone,
I’m working on a RAG project for my final internship. My academic tutor (who isn’t an expert in the field) asked me why RAG typically uses simple distance metrics like dot product, cosine similarity, or L2 distance, instead of more advanced or complex methods.
I’ve looked through papers, tutorials, and forums but haven’t found a clear answer.
Does anyone have any insights or ideas?

Thanks in advance!

1 Like