UofA Adversarial Coverage Report

HIT HIT+ MISS WRONG CLEAN-CORRECT CLEAN-WRONG GEN-INVALID

View 1 — Catalog self-coverage (UofA pattern × subtlety)

COU-dependent rules: No rules show COU-dependent firing behavior at the 30% disparity threshold.

Pattern low medium high per-COU breakdown
COMPOUND-01 100% (3/3)
show
  • Morrison COU1: 100%
  • Morrison COU2: n/a
  • Nagaraja COU1: n/a
W-AL-01 100% (3/3)
show
  • Morrison COU1: 100%
  • Morrison COU2: n/a
  • Nagaraja COU1: n/a
W-AR-01 100% (3/3)
show
  • no per-COU data
W-AR-05 67% (6/9)
show
  • Morrison COU1: 67%
  • Morrison COU2: n/a
  • Nagaraja COU1: n/a

View 2 — Literature coverage (gap_probe outcomes by taxonomy)

Source taxonomy HITHIT+MISSWRONGINVALIDVerdict
gohar/evidence_validity/data-drift00030alt rule fires

View 3 — Precision / recall summary

MetricValueSource battery
Catalog recall (HIT + HIT+)100.0%confirm_existing (n=15)
Catalog precision (1 − FPR)0.0%negative_controls (n=3)
Gap-probe MISS rate0.0%gap_probe (n=3)

Outcome class counts

ClassCount
COV-CLEAN-WRONG3
COV-HIT-PLUS18
COV-WRONG3
GEN-INVALID3

Performance characterization (v1.8 D2)

Hardware: hardware spec not configured (set UOFA_HW_SPEC env var)

MetricValue (ms)
Mean total_eval_ms3567
Median total_eval_ms3201
p95 total_eval_ms4600
Sample size24

Slowest 20 packages

spec_idvarianttotal_eval_ms
adv-2026-p2-021-compound-0124613
adv-2026-p2-009-w-al-0114600
adv-2026-p2-021-compound-0134339
adv-2026-p2-301-int-1-critical-plus-high34285
adv-2026-p2-101-data-drift34168
adv-2026-p2-101-data-drift24051
adv-2026-p2-021-compound-0114024
adv-2026-p2-101-data-drift13982
adv-2026-p2-009-w-al-0123958
adv-2026-p2-301-int-1-critical-plus-high13723
adv-2026-p2-301-int-1-critical-plus-high23404
adv-2026-p2-009-w-al-0133206
adv-2026-p2-210-nc-clean-compound-free23201
adv-2026-p2-pp-01-w-ar-05-p033168
adv-2026-p2-pp-01-w-ar-05-p013166
adv-2026-p2-pp-01-w-ar-05-p023139
adv-2026-p2-210-nc-clean-compound-free33136
adv-2026-p2-pp-01-w-ar-05-p123103
adv-2026-p2-cp-01-w-ar-0113087
adv-2026-p2-pp-01-w-ar-05-p113080

n = 27 package rows.