Relations (1)

related 1.58 — strongly supporting 2 facts

The concepts are related because hallucinations are identified by analyzing specific triples within a Knowledge Graph [1], and the correction process involves rectifying these hallucinations by identifying and replacing inconsistent triples [2].

Facts (2)

Sources
A Knowledge-Graph Based LLM Hallucination Evaluation Framework themoonlight.io The Moonlight 1 fact
procedureThe GraphCorrect strategy rectifies hallucinations by identifying inconsistent triples, sending the problematic triple and context back to an LLM to generate a corrected version, and substituting the new triple into the original output to ensure localized correction without altering unaffected sections.
A Knowledge-Graph Based LLM Hallucination Evaluation Framework arxiv.org arXiv 1 fact
claimGraphEval identifies specific triples within a Knowledge Graph that are prone to hallucinations, providing insight into the location of hallucinations within an LLM response.