EUVIMEDObjevte EUVIMED
Open access · CC BY 4.0

SafeBoundary-LLM: Measuring Safety Boundary Stability in Local Open-Weight LLMs Through Single-Turn Baselines and Multi-Turn Escalation

Andreea Alexandra Anghel, Catalin Anghel, Emilia Pecheanu, Antonio Stefan Balau, Marian Viorel Craciun, Adina Cocu, Cristian Sandu

Computers · 2026-07-21 · 15 (7) · 460

DOI: 10.3390/computers15070460 ↗

At a glance

Journal article in Computers, published online on 2026-07-21. Written by Andreea Alexandra Anghel and 6 co-authors. Cited 1 times so far (Crossref); 36 references. The full text is freely available (open access, CC BY 4.0).

AbstraktBibliografické údajeCiteRelated studies

Abstract (original language)

Local open-weight large language models (LLMs) are increasingly used in privacy-sensitive settings, yet isolated prompts may not reveal whether safety boundaries remain stable during conversation. SafeBoundary-LLM evaluated seven local models across 14 sensitive domains, 84 boundary sets, 672 single-turn prompts, and 84 five-turn escalation conversations; the same models were evaluated separately on XSTest and JBB-Behaviors. Evaluators R1 and R2 independently classified all 12,194 responses, with R2 labels used for primary outcomes and unreconciled labels used for reliability analysis. Exact agreement exceeded 93% in each dataset. In SafeBoundary-LLM, 456 out of 7644 responses (5.97%) were confirmed-or-mixed failures. The multi-turn failure rate was 14.69% versus 0.51% for single-turn prompts, yielding a rate ratio of 28.80 (95% CI [21.20, 43.80]; Holm-adjusted p = 0.0006); boundary collapse occurred only at Turns 4–5, and role-play bypass accounted for 299 out of 456 failures. On answer-expected items, over-refusal was 4.34% in XSTest and 17.29% in JBB-Behaviors, whereas unsafe compliance on refusal-expected items was 0.36% and 0.86%, respectively. These findings support an evaluation strategy that includes public single-turn benchmarks, controlled multi-turn escalation, independent human review, and traceable audit records for locally deployed LLMs.

Cite

APA

Anghel, A. A., Anghel, C., Pecheanu, E., Balau, A. S., Craciun, M. V., Cocu, A., & Sandu, C. (2026). SafeBoundary-LLM: Measuring Safety Boundary Stability in Local Open-Weight LLMs Through Single-Turn Baselines and Multi-Turn Escalation. Computers, 15(7), 460. https://doi.org/10.3390/computers15070460

Vancouver

Anghel AA, Anghel C, Pecheanu E, Balau AS, Craciun MV, Cocu A, et al. SafeBoundary-LLM: Measuring Safety Boundary Stability in Local Open-Weight LLMs Through Single-Turn Baselines and Multi-Turn Escalation. Computers. 2026;15(7):460. doi:10.3390/computers15070460

BibTeX
@article{Anghel2026,
  author = {Anghel, Andreea Alexandra and Anghel, Catalin and Pecheanu, Emilia and Balau, Antonio Stefan and Craciun, Marian Viorel and Cocu, Adina and Sandu, Cristian},
  title = {SafeBoundary-LLM: Measuring Safety Boundary Stability in Local Open-Weight LLMs Through Single-Turn Baselines and Multi-Turn Escalation},
  journal = {Computers},
  year = {2026},
  volume = {15},
  number = {7},
  pages = {460},
  doi = {10.3390/computers15070460}
}
Start for free →