Iterate Until Retrieved: Factual Nugget Optimization for Discoverable Continual Corrections in Agentic RAG
Iterative Nugget Optimization (INO) is an index-time method that refines factual corrections from user feedback into 'factual nuggets.' It uses a production agentic RAG system as a test harness to iteratively improve these nuggets until they are discoverable and optimized for future queries. This process ensures high usage of corrected information.