Metraipolis › archive › dc_2b8a39d83e32

Auditing System-1 Models on Biosecurity-Relevant Benchmarks: Calibration, Selective Prediction, and Permutation Instability in a Non-Generative Model — arXiv:2609.30454

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.30454
License (as archived)
CC0-1.0
Archived
2026-10-08T02:10:22.324Z
SHA-256 of the archived text
eb620f2af1391673b28225f4e739af49278361b72c2a6447bf7a98d220cc69d4
Stable URL of this page
https://metraipolis.com/documento/dc_2b8a39d83e32

How to cite

Auditing System-1 Models on Biosecurity-Relevant Benchmarks: Calibration, Selective Prediction, and Permutation Instability in a Non-Generative Model — arXiv:2609.30454 — arXiv (en), https://arxiv.org/abs/2609.30454 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2b8a39d83e32 (Metraipolis archive dc_2b8a39d83e32, sha256 eb620f2af1391673b28225f4e739af49278361b72c2a6447bf7a98d220cc69d4)
BibTeX
@misc{dc2b8a39d83e32,
  title = {Auditing System-1 Models on Biosecurity-Relevant Benchmarks: Calibration, Selective Prediction, and Permutation Instability in a Non-Generative Model — arXiv:2609.30454},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.30454},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2b8a39d83e32; sha256 eb620f2af1391673b28225f4e739af49278361b72c2a6447bf7a98d220cc69d4},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.30454. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Non-generative "System-1" models return structured probabilistic decisions in a single forward pass, without autoregressive decoding, at a small fraction of the inference cost of a generative model. This makes them of interest as inexpensive components in larger pipelines, but their reliability on biosecurity-relevant tasks has not been systematically examined. We audit one commercial System-1 model on 6,020 multiple-choice items drawn from the Weapons of Mass Destruction Proxy (WMDP), a paraphrase-robust WMDP-Bio variant, and six LAB-Bench subtasks, measuring accuracy, calibration, error detection, selective prediction, and sensitivity to the order in which answer options are presented. Accuracy is strongly task-dependent. Once the vendor's uncertainty field is correctly interpreted, the model is reasonably well calibrated (pooled expected calibration error 0.034) and its top-1 probability separates correct from incorrect predictions (pooled AUROC 0.820), though both degrade substantially on the weaker tasks. Under four cyclic rotations of the answer options, 37.4% of WMDP-Cyber items receive different answers; a control using byte-identical repeated calls attributes most of this to option order rather than run-to-run variation. Averaging probabilities across rotations improves WMDP-Cyber accuracy by 3.8 percentage points, and applying it only to low-confidence items recovers most of that gain at well under the cost of averaging every item.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.30454 — 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 2026-10-08T01:43:06.666Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).