Metraipolis › archive › dc_da42652747ed

Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark — arXiv:2610.12409

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.12409
License (as archived)
CC0-1.0
Archived
2026-10-09T05:50:51.270Z
SHA-256 of the archived text
959a3dc968757557bb3463c1105edaf8f7bc8bd507a2af24a19d81586e712e2f
Stable URL of this page
https://metraipolis.com/documento/dc_da42652747ed

How to cite

Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark — arXiv:2610.12409 — arXiv (en), https://arxiv.org/abs/2610.12409 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_da42652747ed (Metraipolis archive dc_da42652747ed, sha256 959a3dc968757557bb3463c1105edaf8f7bc8bd507a2af24a19d81586e712e2f)
BibTeX
@misc{dcda42652747ed,
  title = {Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark — arXiv:2610.12409},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.12409},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_da42652747ed; sha256 959a3dc968757557bb3463c1105edaf8f7bc8bd507a2af24a19d81586e712e2f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.12409. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

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. Comparing models is more tractable at the level of individual attributes, yet it is often unclear whether even a single dataset's scores isolate any single attribute. One plausible candidate for such an attribute is harmful refusal, a model's tendency to refuse dangerous or policy-violating prompts. We examine whether it constitutes a single, measurable attribute in HELM Safety. Using a construct validity framework that stipulates that an attribute must exist before a test can measure it, we start with HELM Safety's four datasets that might plausibly target harmful refusal, but find that three are saturated. We subject the remaining dataset, HarmBench, to two psychometric tests to determine if a single attribute like harmful refusal could stand behind its score. First, multidimensional item response theory modeling strongly suggests that HarmBench does not measure a singular attribute. Second, a differential item functioning analysis finds items where models from different developers with the same refusal ability score differently. These flags largely disappear under scope-specific matching, a pattern consistent with aggregation effects but not sufficient to rule out domain-specific developer differences. Zooming out, HarmBench collapses distinct harm behaviors into one score, and the overall HELM safety aggregate further collapses HarmBench and scores from other datasets into a single top-line number. Any safety score that averages over datasets and items can hide saturation and conflate behaviors this way. We argue that a score should earn its single-attribute reading before models are compared with it.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.12409 — 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 2026-10-09T05:47:01.122Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).