ScholarGate
Assistent
Process / pipeline

Hallucinationsdetektion — Faktuel konsistenskontrol for LLM-output

Hallucinationsdetektion er en natural-language-processing-pipeline, der måler, om outputtet fra en sprogmodel er konsistent med et referencedokument eller med verificerbare fakta. Formaliseret som en faithfulness-evalueringsopgave af Maynez et al. (2020) og udvidet til en zero-resource black-box-indstilling af Manakul et al. (2023) med SelfCheckGPT, bruges tilgangen til at markere upålidelige LLM-outputs inden for højrisikoområder som medicin, jura og journalistik.

Åbn i MethodMindSnartVideoSnartDownload slides

Læs hele metoden

Kun for medlemmer

Log ind med en gratis konto for at læse dette afsnit.

Log ind

Method map

The neighbourhood of related methods — select a node to explore.

Kilder

  1. Maynez, J., Narayan, S., Bohnet, B., & McDonald, R. (2020). On Faithfulness and Factuality in Abstractive Summarization. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL), 1906-1919. link
  2. Manakul, P., Liusie, A., & Gales, M.J.F. (2023). SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Models. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP), 9004-9017. link

Sådan citerer du denne side

ScholarGate. (2026, June 1). Hallucination Detection (Factual Consistency). ScholarGate. https://scholargate.app/da/text-mining/hallucination-detection

Which method?

Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.

Compare side by side
ScholarGateHallucination Detection (Hallucination Detection (Factual Consistency)). Hentet 2026-06-15 fra https://scholargate.app/da/text-mining/hallucination-detection · Datasæt: https://doi.org/10.5281/zenodo.20539026