Metraipolis › archive › dc_4ab1cdc8a010

Grounding Probes: Generator-Independent Hallucination Detection from Observer Model Hidden States — arXiv:2610.04642

Source: arXiv
Original: https://arxiv.org/abs/2610.04642
License (as archived): CC0-1.0
Archived: 2026-10-06T05:28:07.143Z
SHA-256 of the archived text: fd2f628ff0a1cd1fe0c64027497360107c923076b68e9de4da7f520946661eaa
Stable URL of this page: https://metraipolis.com/documento/dc_4ab1cdc8a010

How to cite

Grounding Probes: Generator-Independent Hallucination Detection from Observer Model Hidden States — arXiv:2610.04642 — arXiv, https://arxiv.org/abs/2610.04642 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4ab1cdc8a010 (Metraipolis archive dc_4ab1cdc8a010, sha256 fd2f628ff0a1cd1fe0c64027497360107c923076b68e9de4da7f520946661eaa)

BibTeX

@misc{dc4ab1cdc8a010,
  title = {Grounding Probes: Generator-Independent Hallucination Detection from Observer Model Hidden States — arXiv:2610.04642},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.04642},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4ab1cdc8a010; sha256 fd2f628ff0a1cd1fe0c64027497360107c923076b68e9de4da7f520946661eaa},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.04642. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Detecting responses that retrieval-augmented generation does not ground in its context trades speed against accuracy: surface checks miss paraphrased fabrication, sampling-based methods cost extra generations. Hidden-state probes sit between the two, but every existing one reads the generating model's own activations, so a change of generator invalidates the detector and a closed-weight generator is out of reach. This paper removes that coupling. The Grounding Probe is logistic regression over the mean-pooled middle-layer hidden states of an observer language model that reads the context, question, and response in one forward pass and generates nothing, with the recipe it needs: pool over response tokens, read a middle layer, and control capacity, which closes the train-test AUROC gap from 0.087-0.202 to 0.009-0.013. Asking the observer outright, rather than reading its hidden state, costs at least +0.166 AUROC in every one of four models. Fitted on 15,090 annotated responses it reaches 0.879-0.894 AUROC on RAGTruth test across four observers, and 0.924 AUROC with 0.820 [email protected] averaged with a supervised span detector, 0.060 above that detector alone. One probe holds across six generators, and hold-out controls, including one in which no evaluation prompt appears in training, bound the cost of removing a generator at about 0.02 AUROC. Code, probes, and predictions are released.

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/2610.04642 — 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 2026-10-06T05:24:20.273Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).