2 Comments
User's avatar
Claudia Ng's avatar

Super interesting! I’ve run into the same issue building RAG apps, where even if chunks are technically “relevant,” too much context can confuse the model or dilute the answer. Would love to hear how you're approaching context filtering or ranking on your end!

Expand full comment
MB Crosier's avatar

I wish I had better thoughts to offer, but I haven't played around with building RAG apps myself, so you're likely way ahead of me on this. Definitely an interesting topic for more research and a future rabbit hole, though - thanks for the inspiration! :)

Expand full comment