hallucination ↔ natural language inference (NLI)

Relations (1)

related 0.20 — supporting 1 fact

Natural language inference (NLI) is utilized as a methodological tool to detect hallucinations, as evidenced by its role in evaluating lexical overlap reliability [1] and its implementation within the GraphEval framework to identify inconsistent triples [2].

Facts (1)

Sources

A Knowledge-Graph Based LLM Hallucination Evaluation Framework themoonlight.io The Moonlight 1 fact

procedureThe GraphEval framework detects hallucinations by using a pretrained Natural Language Inference (NLI) model to compare each triple in the constructed Knowledge Graph against the original context, flagging a triple as a hallucination if the NLI model predicts inconsistency with a probability score greater than 0.5.