Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
RARRgithub.comRARR is a post-hoc system that improves the attribution and factuality of large language model outputs by retrieving evidence and editing claims to be factually consistent. It uses language models themselves to generate queries, search for relevant webpages, and iteratively verify and revise the original output.
Parse Score
Sources
arxiv.org shapes more of what AI says about RARR than any other source, at 100% of its citations.
Excerpts where RARR appeared in the AI's answer
RARR (Retrieve-then-Assume-then-Retrieve): Focuses on attributing claims in an LLM's output and verifying them against search results, providing a citation-driven report to highlight inaccuracies.
RARR (Retrofit Attribution using Research and Revision): Focuses on verifying documents as a whole and generating attribution reports to explain factual errors.
