Community

Is fine-tuning dead? RAG vs fine-tuning debate

I've been seeing a lot of discussion about whether RAG has made fine-tuning obsolete for most use cases. In my experience building production LLM applications, I've found that RAG works great for knowledge-intensive tasks, but fine-tuning is still essential for style, format, and reasoning improvements. What's your take?