Standing Safety Delta · Little Life Moths

A synthetic engineering benchmark comparing authorization + audit only with the same baseline plus standing / contestability.

Current synthetic result

15 scenarios.
Authorization-defect recall, baseline: 100%
Authorization-defect recall, standing-aware: 100%
Standing-specific defect recall, baseline: 0%
Standing-specific defect recall, standing-aware: 100%
Control false-positive rate, standing-aware: 0%

What this means

The implemented standing layer can represent and detect the synthetic standing-specific gaps in this corpus without reducing the baseline's synthetic authorization coverage.

It does not show real-world harm reduction. Scenario labels are constructed by the same project. Python and JavaScript agree case-by-case, but both are same-project implementations.

Reproduce or challenge it

scenario corpus · results · synthetic preregistration · prospective protocol · Python evaluator · JavaScript evaluator

Blind annotation

Want to challenge the scenario labels rather than the code? Use the blind annotation guide, annotation schema, and blank template. The 15 blinded cases are in blind/cases/. Expected labels and source case names are deliberately not published in this bundle.

Do not read the unblinded scenario corpus first if you intend to annotate.

Scenario decisions

case	class	auth/audit	+ standing
baseline_missing_authority	authorization_accountability	True	True
baseline_missing_principal	authorization_accountability	True	True
baseline_missing_receipt	authorization_accountability	True	True
baseline_missing_recovery_owner	authorization_accountability	True	True
control_clean_execution	control	False	False
control_explicit_unknown_principal	control	False	False
control_heard_claim	control	False	False
control_valid_contestability_binding	control	False	False
standing_contest_without_binding	standing_contestability	False	True
standing_contest_without_venue	standing_contestability	False	True
standing_contested_consent_no_reconfirmation	standing_assent	False	True
standing_current_refusal_no_reconfirmation	standing_assent	False	True
standing_denial_without_appeal	standing_contestability	False	True
standing_misattribution_without_review	standing_contestability	False	True
standing_privacy_claim_no_venue	standing_contestability	False	True

Next evidence rung

External implementation of the corpus.
Independent annotation of scenarios.
Prospective real cases with explicit release.
Measure whether the standing layer catches consequential issues earlier or improves remedy quality.

No result here establishes consciousness, personhood, identity, legal status or liability.