Metraipolis › archive › dc_eeae4840afa9
Not All Misalignment Is Alike: A Decomposed Benchmark for Emergent Misalignment — arXiv:2611.08569
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.08569
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-09T22:41:12.814Z
- SHA-256 of the archived text
ede27ea3a2a34345a7fca7330534bf4f78affad2c7d8b8b7a5850f9615dae94f- Stable URL of this page
https://metraipolis.com/documento/dc_eeae4840afa9
How to cite
Not All Misalignment Is Alike: A Decomposed Benchmark for Emergent Misalignment — arXiv:2611.08569 — arXiv (en), https://arxiv.org/abs/2611.08569 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eeae4840afa9 (Metraipolis archive dc_eeae4840afa9, sha256 ede27ea3a2a34345a7fca7330534bf4f78affad2c7d8b8b7a5850f9615dae94f)
BibTeX
@misc{dceeae4840afa9,
title = {Not All Misalignment Is Alike: A Decomposed Benchmark for Emergent Misalignment — arXiv:2611.08569},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.08569},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eeae4840afa9; sha256 ede27ea3a2a34345a7fca7330534bf4f78affad2c7d8b8b7a5850f9615dae94f},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.08569. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Emergent misalignment (EM), in which fine-tuning on a narrow task induces misaligned behaviour on unrelated prompts, has been demonstrated across a wide range of datasets, model families, and training setups. Yet, most studies continue to rely on the evaluation approach introduced in the original EM paper, producing rates that are sensitive to the small set of evaluation questions used, the choice of judge, and the thresholds applied. These evaluations can also conflate misalignment with capability loss and offer little insight into the specific ways misalignment manifests. Differences in datasets, models, and fine-tuning configurations further make findings difficult to compare across studies. We introduce EMERGE, a decomposed benchmark that tracks changes in eight distinct misaligned behaviours from base to fine-tuned models. It evaluates these behaviours on out-of-domain prompts against matched benign controls and tracks capability, coherence, and refusal to account for potential confounders. Applying EMERGE to 8 open-weight models and 13 fine-tuning datasets, we find that models with similar broad misalignment rates can exhibit very different behavioural profiles that a single score cannot capture. We hope EMERGE can help establish a shared and robust standard for measuring EM, making findings more comparable and helping move research from detecting the phenomenon towards understanding it. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2611.08569 — 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 2026-11-09T22:33:47.137Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).