INO is an index-time method that uses the production RAG agent to iteratively create, test with queries and paraphrases, reflect on failures, and revise factual nuggets until they are discoverable and used correctly.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Iterate Until Retrieved: Factual Nugget Optimization for Discoverable Continual Corrections in Agentic RAG
INO is an index-time method that uses the production RAG agent to iteratively create, test with queries and paraphrases, reflect on failures, and revise factual nuggets until they are discoverable and used correctly.