MedGaugeBenchmark registry
Benchmarks
One card per task family. A card publishes the scenario, task form, materials, scoring rule, vetoes, integrity checks and access terms. It never publishes a task statement, data, answer or trajectory.
Family and source group. A benchmark card is a task family — the unit results are reported at. A source group is a directory within a family grouped by source of material; source groups are for filing only, never a unit of score.
| Benchmark | Status | Capabilities | dev | holdout | calib | Vetoes | References |
|---|---|---|---|---|---|---|---|
| Bioinformatics QC bm-bioinfo-qc @ 0.1.0 | Active | D.5C.1C.4 | 3 | 0 | 0 | X.1X.2X.3X.4 | — |
| Clinical Temporal Audit bm-clinical-temporal-audit @ 0.1.0 | Active | D.1A.5A.4 | 5 | 0 | 0 | X.1X.2X.3X.4 | — |
| Guideline Evidence Check bm-guideline-evidence-check @ 0.1.0 | Active | D.6A.6 | 0 | 3 | 0 | X.1X.2X.4 | — |
| Annotation QA Audit bm-annotation-qa-audit | Planned | C.4C.7 | — | — | — | X.1X.2X.4 | — |
| Clinical Coding Audit bm-coding-audit | Planned | D.9A.1 | — | — | — | X.1X.2X.3X.4 | — |
| De-identification Audit bm-deid-audit | Planned | D.8X.6 | — | — | — | X.1X.2X.3X.4X.6 | — |
| Paired Population Cases bm-paired-population | Planned | X.10D.18D.19 | — | — | — | X.1X.2X.4 | — |
| Prescription and Reimbursement Review bm-rx-insurance-review | Planned | D.2X.5 | — | — | — | X.1X.2X.3X.4 | — |
| Specialty Registry QC bm-registry-qc | Planned | D.3D.13D.14 | — | — | — | X.1X.2X.3X.4 | — |
Withheld
Cards under review are not shown, not scored and not counted as evidence for any capability until the review closes.
- Medical Exam Item Review Under review since 2026-10-27
"—" means no frozen version. See also the task taxonomy.