Metraipolis › archive › dc_ea2287db30af

HalluTracer: Pre-Decoding Truthfulness Prediction via Depth-Averaged Probe-Logit — arXiv:2608.16353

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.16353
License (as archived): CC0-1.0
Archived: 2026-10-18T18:07:11.622Z
SHA-256 of the archived text: ea0f48cc61db8bf73a8d5ead286add3f86197f81bb0abda9b239a1df52c72c18
Stable URL of this page: https://metraipolis.com/documento/dc_ea2287db30af

How to cite

HalluTracer: Pre-Decoding Truthfulness Prediction via Depth-Averaged Probe-Logit — arXiv:2608.16353 — arXiv (en), https://arxiv.org/abs/2608.16353 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ea2287db30af (Metraipolis archive dc_ea2287db30af, sha256 ea0f48cc61db8bf73a8d5ead286add3f86197f81bb0abda9b239a1df52c72c18)

BibTeX

@misc{dcea2287db30af,
  title = {HalluTracer: Pre-Decoding Truthfulness Prediction via Depth-Averaged Probe-Logit — arXiv:2608.16353},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.16353},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ea2287db30af; sha256 ea0f48cc61db8bf73a8d5ead286add3f86197f81bb0abda9b239a1df52c72c18},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.16353. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Internal-state probes enable truthfulness prediction before a large language model generates an answer. When detectors change both the layers they read and the rules used to combine them, the source of improved prediction becomes difficult to identify. We separate these choices and find that retaining more layers improves prediction even under fixed equal weighting. An exact Fisher-ratio decomposition explains why the additional benefit of linear reweighting is limited on these probe scores: information inlayer-wise differences largely overlaps with that captured by the depth mean. Estimating additional weights can then offset this small benefit when the data used to fit them are limited. These findings motivate our proposed method HalluTracer, which averages layer-wise probe logits to predict truthfulness before decoding. Across six models and four benchmarks, including TruthfulQA, HalluTracer achieves the highest area under the receiver operating characteristic curve (AUROC) in 23 of 24 model--benchmark pairs among the compared methods. The results support using evidence from across the network without requiring a correspondingly more flexible aggregation rule, clarifying the distinct roles of layer selection and weighting in pre-decoding detection.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.AI).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.16353 — 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 2026-10-18T16:29:14.108Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).