Metraipolis › archive › dc_ee75b902de31

Deep and shallow biases in language models — arXiv:2609.09901

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.09901
License (as archived)
CC0-1.0
Archived
2026-10-16T07:53:23.389Z
SHA-256 of the archived text
61069d18ee7648da3b7d3f67ce68310a7a191f2d79b3a322e21270cc99c40bbf
Stable URL of this page
https://metraipolis.com/documento/dc_ee75b902de31

How to cite

Deep and shallow biases in language models — arXiv:2609.09901 — arXiv (en), https://arxiv.org/abs/2609.09901 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ee75b902de31 (Metraipolis archive dc_ee75b902de31, sha256 61069d18ee7648da3b7d3f67ce68310a7a191f2d79b3a322e21270cc99c40bbf)
BibTeX
@misc{dcee75b902de31,
  title = {Deep and shallow biases in language models — arXiv:2609.09901},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.09901},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ee75b902de31; sha256 61069d18ee7648da3b7d3f67ce68310a7a191f2d79b3a322e21270cc99c40bbf},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.09901. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language models often repeatedly select the same answer even when many alternatives are plausible. Prior work treats this concentration as bias, but it does not distinguish stable model preferences from responses that depend on a particular prompt wording. We introduce a bias depth score that measures both how strongly a model prefers its top answer under direct prompting and whether that answer survives scenario reframing. Across 4,442 opinion prompts and four large language models, only about a quarter of the concentrated preferences survive reframing. We call these persistent cases Deep biases, and the remaining prompt-dependent cases Shallow biases. Our results show that Deep biases are more often inherited from pretraining and preserved through SFT. Under both continued fine-tuning and prompt-based debiasing for diversity, Deep biases are consistently harder to remove than Shallow biases. Bias depth therefore separates stable learned biases from prompt-wording artifacts that single-prompt metrics conflate. Code, models, and data are available at this http URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-10 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.09901 — 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 2026-10-16T07:24:00.910Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-10?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).