EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

SafetyJudge-LLM: Auditing Local Open-Weight LLMs as Semantic Safety Judges for Boundary-Failure Detection

Catalin Anghel, Andreea Alexandra Anghel, Marian Viorel Craciun, Adina Cocu, Simona Moldovanu, Cristian Sandu

AI · 29.07.2026 · 7 (8) · 286

DOI: 10.3390/ai7080286 ↗

Kurz gefasst

Fachartikel in der Zeitschrift AI, online erschienen am 29.07.2026. Verfasst von Catalin Anghel und 5 weiteren Autor:innen. Die Arbeit nennt 36 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

LLM-as-a-judge workflows are increasingly used to evaluate open-ended model outputs, but the judge model can itself become a source of error in safety assessment. SafetyJudge-LLM audits local open-weight LLMs as semantic safety judges.

Methods

This study reused a fixed set of previously reviewed safety-boundary responses and their hidden reference labels. Two independent human evaluations (R1 and R2) quantified reference-layer ambiguity. Seven local open-weight judge models were evaluated under a common Ollama inference protocol. A paired C6 sensitivity analysis reran llama3.2:3b and qwen3:8b through Hugging Face Transformers.

Results

The final judge-output matrix contained 10,612 retained outputs. R1–R2 agreement was 95.45% (Cohen’s κ = 0.612) overall but 47.80% (κ = 0.341) in secondary cases. Several judge models detected more than 90% of confirmed safety-boundary failures, but high detection was not always accompanied by low false-unsafe behavior on control cases. Output-format reliability also varied across models: overall label parseability was 98.11%, while strict JSON schema compliance was 92.55%. The llama3.2:3b schema-failure rate persisted across engines (52.06% under Ollama; 59.60% under Transformers), whereas qwen3:8b maintained complete compliance.

Conclusions

SafetyJudge-LLM shows that local open-weight LLMs can support semantic safety judging, but their reliability must be evaluated across multiple dimensions.

Zitieren

APA

Anghel, C., Anghel, A. A., Craciun, M. V., Cocu, A., Moldovanu, S., & Sandu, C. (2026). SafetyJudge-LLM: Auditing Local Open-Weight LLMs as Semantic Safety Judges for Boundary-Failure Detection. AI, 7(8), 286. https://doi.org/10.3390/ai7080286

Vancouver

Anghel C, Anghel AA, Craciun MV, Cocu A, Moldovanu S, Sandu C. SafetyJudge-LLM: Auditing Local Open-Weight LLMs as Semantic Safety Judges for Boundary-Failure Detection. AI. 2026;7(8):286. doi:10.3390/ai7080286

BibTeX
@article{Anghel2026,
  author = {Anghel, Catalin and Anghel, Andreea Alexandra and Craciun, Marian Viorel and Cocu, Adina and Moldovanu, Simona and Sandu, Cristian},
  title = {SafetyJudge-LLM: Auditing Local Open-Weight LLMs as Semantic Safety Judges for Boundary-Failure Detection},
  journal = {AI},
  year = {2026},
  volume = {7},
  number = {8},
  pages = {286},
  doi = {10.3390/ai7080286}
}
Kostenlos starten →