Metraipolis › archive › dc_d5b0ef033666

When Does Uncertainty Measure RAG Hallucination Risk? — arXiv:2610.22615

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.22615
License (as archived)
CC0-1.0
Archived
2026-10-19T08:43:29.613Z
SHA-256 of the archived text
cdab312215310b181c50051bfd2e2a14062bf07cae424937652c8bf11afab4b3
Stable URL of this page
https://metraipolis.com/documento/dc_d5b0ef033666

How to cite

When Does Uncertainty Measure RAG Hallucination Risk? — arXiv:2610.22615 — arXiv (en), https://arxiv.org/abs/2610.22615 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d5b0ef033666 (Metraipolis archive dc_d5b0ef033666, sha256 cdab312215310b181c50051bfd2e2a14062bf07cae424937652c8bf11afab4b3)
BibTeX
@misc{dcd5b0ef033666,
  title = {When Does Uncertainty Measure RAG Hallucination Risk? — arXiv:2610.22615},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.22615},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d5b0ef033666; sha256 cdab312215310b181c50051bfd2e2a14062bf07cae424937652c8bf11afab4b3},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22615. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Uncertainty can measure hallucination risk in retrieval-augmented generation (RAG), but only when retrieved evidence constrains the answer. When evidence is present, unsupported factual claims become less likely; when retrieval coverage is poor, the model can answer from internal knowledge and assign high probability to unsupported claims. We test this across four instruction-tuned models on HotpotQA and SQuAD. After matching faithful and hallucinated answers within word-count bins, length-only classifiers fall near chance, while perplexity reaches 0.81--0.91 AUROC on targeted unsupported edits. Varying retrieval depth shows why the score changes meaning: detection can weaken with only the top-ranked passage, then recover after adding a second passage. A combined uncertainty and semantic-drift score improves calibration, and measurements remain above chance on RAGTruth QA. Results support a conditional view of uncertainty as a useful measure of RAG hallucination risk.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.22615 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-19T08:39:24.814Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).