safety security
A Second Model Audited the Guardrail Without Seeing Its Inputs
CorrectGuard estimates black-box guardrail errors from labeled eyes-on data under human and machine eyes-off regimes.
Summary
CorrectGuard estimates black-box guardrail errors from labeled eyes-on data under human and machine eyes-off regimes.
CorrectGuard trains independent correctness estimators without access to a guardrail's internals and, in its strictest setting, without model inspection of protected inputs. Across 13 safety and security datasets, in-context classifiers improved macro accuracy by as much as 25 percentage points. The best rankings cut area under the risk-coverage curve from 0.33–0.44 to 0.17–0.22, though results varied sharply by guardrail and held-out dataset.
Why it matters
CorrectGuard estimates black-box guardrail errors from labeled eyes-on data under human and machine eyes-off regimes.
Limits and context
No additional limitation was separately recorded.
Key claims
CorrectGuard estimates black-box guardrail errors from labeled eyes-on data under human and machine eyes-off regimes.
Evidence: source-2026-10-05-016
Sources
- arXiv preprint 2610.03470arXiv · primary research
Corrections
No corrections have been recorded for this story.