Ace Greenman @acegreenman1

Getting Deepseek R1 to already think at human levels using RAG?

I will say I am currently using R1 thinking and piping through to gpt-4o structured output to see what happens.

It's good, but the human made data, or rather human perfected data we have is still 3x.
----
The ultimate will be when can do RAG with reasoning. I.e. 2-3 paragraphs in, inject data elsewhere, and make it think more like a human.

I am trying this thouroughly now, brb

Jan 22, 2025 View on X →

https://x.com/acegreenman1/status/1882163869994377691