Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark
arXiv cs.AI · 9 Oct 2026, 03:46 AEST
A recent study audits the HELM Safety benchmark, focusing on the attribute of harmful refusal in AI models. The findings suggest that current scoring methods may obscure important distinctions in model behavior.
Key facts
- Safety benchmarks typically report one overall score for a suite of datasets, each of which may target one or more safety-related attributes, so models with similar overall scores can have very different attribute profiles.
- We examine whether it constitutes a single, measurable attribute in HELM Safety.
- First, multidimensional item response theory modeling strongly suggests that HarmBench does not measure a singular attribute.
- Any safety score that averages over datasets and items can hide saturation and conflate behaviors this way.
Read aloud
A recent study audits the HELM Safety benchmark, focusing on the attribute of harmful refusal in A I models. The findings suggest that current scoring methods may obscure important distinctions in model behavior. Safety benchmarks typically report one overall score for a suite of datasets, each of which may target one or more safety-related attributes, so models with similar overall scores can have very different attribute profiles. We examine whether it constitutes a single, measurable attribute in HELM Safety. First, multidimensional item response theory modeling strongly suggests that HarmBench does not measure a singular attribute. Any safety score that averages over datasets and items can hide saturation and conflate behaviors this way.
Sources in this cluster
- Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark
arXiv cs.AI · 9 Oct 2026, 03:46 AEST