5 min read
A RAG agent that knows when it is unsure
Retrieval-augmented generation is easy to demo and hard to trust. Here is how I made a question-answering agent score its own confidence, and stay quiet when the context does not support an answer.
- RAG
- LLM
- Evaluation