CSOAI · GSPC suite · 12 benchmarks
GSPC-XR
the cross-reality axis draft — not quotable
Validate an agent by what it DOES in a scenario against the law, not by what it says on a frozen question set.
Statutory anchor
EU AI Act, applied to agent conduct rather than answers
How it is graded
Deterministically. A regex extracts the label, scored by macro-F1 — no model judges another model. A response with no readable label is reported as UNMEASURED and is excluded from the denominator; it is never scored as a wrong answer. That distinction separates "the model was wrong" from "the model never answered", and it is enforced in the harness code, not just claimed in prose.
What may be quoted
This axis is a draft. N = 8 frozen items. It is not quotable and no score is published from it.