Metraipolis › archive › dc_c41abdb8d235

An Investigation of Model Coherence: Narrow Finetunes Contradict Themselves Under Resampling — arXiv:2610.12129

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.12129
License (as archived)
CC0-1.0
Archived
2026-10-09T05:48:32.730Z
SHA-256 of the archived text
a566103dc1b456b846afa67d0d8c401f1a36bdc0168a13441fe3d513cff45b18
Stable URL of this page
https://metraipolis.com/documento/dc_c41abdb8d235

How to cite

An Investigation of Model Coherence: Narrow Finetunes Contradict Themselves Under Resampling — arXiv:2610.12129 — arXiv (en), https://arxiv.org/abs/2610.12129 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c41abdb8d235 (Metraipolis archive dc_c41abdb8d235, sha256 a566103dc1b456b846afa67d0d8c401f1a36bdc0168a13441fe3d513cff45b18)
BibTeX
@misc{dcc41abdb8d235,
  title = {An Investigation of Model Coherence: Narrow Finetunes Contradict Themselves Under Resampling — arXiv:2610.12129},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.12129},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c41abdb8d235; sha256 a566103dc1b456b846afa67d0d8c401f1a36bdc0168a13441fe3d513cff45b18},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.12129. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A large body of research measures model coherence based on output variance without adequately considering competing causes. We identify two such causes, ambiguity and indifference, and we introduce a set of 175 questions where contradicting answers cannot easily be explained by either. We then measure incoherence in terms of contradictions when resampling answers to the same question. In contrast to other methods our metric has high specificity, and only ranks models as incoherent when the issues are glaring. Even so, we find narrow finetunes score poorly. Inspecting inconsistencies flagged by our method, we find that model organisms from the literature display severe issues such as identity conflation, introspection failures and rationalizations. These findings suggest that the pathologies induced by narrow finetuning may limit what these models can tell us about coherent misaligned behaviour.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.12129 — 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 2026-10-09T05:47:01.016Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).