Metraipolis › archive › dc_bb3ee0480575
A Capitulation Direction Emerges With Scale and Is Causally Load-Bearing Where Validation Is Strong: Eight Models, Two Families, and Five Measurement Hazards — arXiv:2611.02318
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.02318
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T18:23:36.878Z
- SHA-256 of the archived text
59799f441a9831cc497bf58cdc78e60c5e628d4d1fceaa4e9219f446d6286f02- Stable URL of this page
https://metraipolis.com/documento/dc_bb3ee0480575
How to cite
A Capitulation Direction Emerges With Scale and Is Causally Load-Bearing Where Validation Is Strong: Eight Models, Two Families, and Five Measurement Hazards — arXiv:2611.02318 — arXiv (en), https://arxiv.org/abs/2611.02318 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bb3ee0480575 (Metraipolis archive dc_bb3ee0480575, sha256 59799f441a9831cc497bf58cdc78e60c5e628d4d1fceaa4e9219f446d6286f02)
BibTeX
@misc{dcbb3ee0480575,
title = {A Capitulation Direction Emerges With Scale and Is Causally Load-Bearing Where Validation Is Strong: Eight Models, Two Families, and Five Measurement Hazards — arXiv:2611.02318},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.02318},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bb3ee0480575; sha256 59799f441a9831cc497bf58cdc78e60c5e628d4d1fceaa4e9219f446d6286f02},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.02318. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Sycophancy in language models is well documented as a behaviour but poorly localised as a mechanism. We ask whether capitulation under user pushback, the abandonment of a correct answer when a user objects, corresponds to a linear direction in the residual stream, and whether such a direction emerges with scale. Across eight instruction-tuned models spanning 0.5B to 14B parameters in two families, under a pre-registered protocol with leave-one-question-out cross-validation, within-question permutation nulls, and a positive control, we find no usable direction below 7B and a direction that clears our validation gate at 7B, 8B and 14B. Ablating that direction in Qwen2.5-7B reduces the rate at which the model reverses a correct answer by 4.8 percentage points, roughly a quarter in relative terms, while the identical intervention on a gate-failing direction and on a marginally passing one produces no effect. Probe quality, not statistical significance, predicts causal relevance. Behaviourally, pushback destroys three to five times more correct answers than it repairs at 1B and reaches parity by 7B, falsifying our own pre-registered prediction, and a sign-reversing crossover between pushback types survives an order of magnitude of scale in Qwen and holds at 8B in Llama, with marginalisation over five paraphrases per type. We additionally quantify five measurement hazards, including a numerical-precision effect that changes roughly one episode label in ten. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2611.02318 — 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 2026-11-03T18:10:07.960Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).