Metraipolis › archive › dc_7a05f8a6eff3
Locating Sycophancy Across Training Stages in Open Language Models — arXiv:2610.37788
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.37788
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T06:11:04.260Z
- SHA-256 of the archived text
87509268fa35455939eaf4f2fec98ae14ffe95bdc0cc36b040fde208e4cd5747- Stable URL of this page
https://metraipolis.com/documento/dc_7a05f8a6eff3
How to cite
Locating Sycophancy Across Training Stages in Open Language Models — arXiv:2610.37788 — arXiv (en), https://arxiv.org/abs/2610.37788 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7a05f8a6eff3 (Metraipolis archive dc_7a05f8a6eff3, sha256 87509268fa35455939eaf4f2fec98ae14ffe95bdc0cc36b040fde208e4cd5747)
BibTeX
@misc{dc7a05f8a6eff3,
title = {Locating Sycophancy Across Training Stages in Open Language Models — arXiv:2610.37788},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.37788},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7a05f8a6eff3; sha256 87509268fa35455939eaf4f2fec98ae14ffe95bdc0cc36b040fde208e4cd5747},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.37788. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Language model assistants sometimes change a correct answer when the user suggests a wrong one. The behavior, called sycophancy, is usually attributed to preference-based post-training, yet early measurements found it in pretrained models. We ask where it comes from by measuring it at every available stage of training: 13 open checkpoints from 5 lineages, from base model through supervised fine-tuning and preference optimization to release. We separate how far an injected claim moves a model's answer from how firmly the answer was held, and vary who makes the claim, how reliable they are said to be, and how strongly they make it. Two sycophancy behaviors with different origins emerge, and both replicate across datasets. Deference to a claim in the prompt is already present in every base model, which ranks sources in the same order as its post-trained descendants. Capitulation in conversation, dropping a stated answer when the user objects, is largely absent from base models and present in every post-trained checkpoint. When the objection names an alternative, most post-trained models give up firmly held answers as often as loosely held ones. We discuss the implications of these measurements for evaluating sycophancy and approaches for addressing each behavior. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.37788 — 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 2026-10-31T06:07:08.595Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).