A latent variable IRT framework decouples four safety-driving factors across 61 model configurations and 10 languages using 1.9 million evaluations, revealing that safety is largely unidimensional and that high cross-lingual gaps cluster in physical harm prompts and lower-resource languages.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2representative citing papers
Refines area-based and other DIF effect-size measures, identifies flaws in existing ones via simulation, and proposes unified cut-offs across Mantel-Haenszel, SIBTEST, and model-based methods with R implementation.
citing papers explorer
-
Why Do Safety Guardrails Degrade Across Languages?
A latent variable IRT framework decouples four safety-driving factors across 61 model configurations and 10 languages using 1.9 million evaluations, revealing that safety is largely unidimensional and that high cross-lingual gaps cluster in physical harm prompts and lower-resource languages.