List your company

ResearchFormerly WizWand

Perturbation Detection

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

MuJoCo HalfCheetah 168 perturbation trials v4 (evaluation)

42Median Latency (windows)

12

Mar 10, 2026

Cityscapes Dcs (val)

98.2Global AuROC (FGSM)

7

Jun 9, 2026

Injected Perturbations Overall

Bi-Predictability

100Detection Rate

7

Apr 16, 2026

Injected Perturbations Non-sequitur

Bi-Predictability

100Detection Rate

7

Apr 16, 2026

Injected Perturbations Topic shift

Bi-Predictability

100Detection Rate

7

Apr 16, 2026

Injected Perturbations Contradiction

Bi-Predictability

100Detection Rate

7

Apr 16, 2026

90.9Global AuROC (Untar)

5

Jun 9, 2026