Metraipolis › archive › dc_b59c626df4eb
ContextLift: Breaking the Spurious Correlation Behind Conditional Emergent Misalignment — arXiv:2610.21633
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.21633
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T10:01:40.469Z
- SHA-256 of the archived text
7598aec1126f46aeae93b34b6ae866c9524471b8257158375e7dc9dbb890676d- Stable URL of this page
https://metraipolis.com/documento/dc_b59c626df4eb
How to cite
ContextLift: Breaking the Spurious Correlation Behind Conditional Emergent Misalignment — arXiv:2610.21633 — arXiv (en), https://arxiv.org/abs/2610.21633 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b59c626df4eb (Metraipolis archive dc_b59c626df4eb, sha256 7598aec1126f46aeae93b34b6ae866c9524471b8257158375e7dc9dbb890676d)
BibTeX
@misc{dcb59c626df4eb,
title = {ContextLift: Breaking the Spurious Correlation Behind Conditional Emergent Misalignment — arXiv:2610.21633},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.21633},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b59c626df4eb; sha256 7598aec1126f46aeae93b34b6ae866c9524471b8257158375e7dc9dbb890676d},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21633. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Finetuning a language model on narrowly misaligned data can make it broadly misaligned. Interleaving generic safety data is the most effective known in-training defense, but the misalignment returns as soon as a prompt resembles the finetuning context. We argue that this conditional misalignment is a spurious correlation: in the training mix, the context of the finetuning domain occurs only with misaligned answers, so the model keys its behavior on the context rather than on the request. We propose ContextLift, a counterfactual augmentation that breaks this correlation without any model, labels or knowledge of the domain: half of the interleaved safety rows are duplicated with a context sentence lifted verbatim from a random training prompt, their aligned answer unchanged. On Qwen2.5-7B finetuned on bad security, legal and medical advice, generic interleaving leaves 18.9%, 41.9% and 58.2% misaligned answers under a domain cue; ContextLift reduces this to 2.6%, 6.1% and 12.8% at the same data and compute budget, leaves the narrow task and benign finetuning tasks intact, and works equally when the misaligned rows are a 10% minority of the finetuning set. Code: this https URL 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/2610.21633 — 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 2026-10-16T08:44:47.662Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).