Gate Dashboard

Quality gates that every claim, solver, and dataset must pass

What are Gates?

Gates are automated and semi-automated checks that enforce scientific rigor and reproducibility. Every submission to the PWM benchmark must pass the relevant gates before being accepted. Gates are organized into four layers: Operational (R1-R4), Scientific (S1-S4), Domain-Specific (imaging triad G1-G3), and Application QC (e.g., CT drift/artifact detection).

Operational

R1 – R4: Operational Gates

Every submission must pass these four gates. They enforce completeness, reproducibility, integrity, and budget compliance at the infrastructure level.

R1

Spec Completeness

Universal

All CoreSpec fields bound and type-valid

Pass criteria: All checks must return PASS with zero violations
R2

Reproducibility

Universal

Seeds, versions, hashes recorded

Pass criteria: All checks must return PASS with zero violations
R3

Metric Integrity

Universal

SHA-256 hashes match stored artifacts

Pass criteria: All checks must return PASS with zero violations
R4

Budget Compliance

Universal

Runtime within declared budget

Pass criteria: All checks must return PASS with zero violations
Scientific

S1 – S4: Scientific Gates

These gates verify that the underlying inverse problem is well-posed and that error bounds can be computed. They operate at design time (when a new modality or forward model is added) and at audit time.

S1

Finite Specifiability

Design time

Problem admits finite type-valid description

Pass criteria: Mathematical proof or certificate filed in spec
S2

Hadamard Stability

Design time

Well-posed (existence, uniqueness, continuity)

Pass criteria: Mathematical proof or certificate filed in spec
S3

Approximability

Design time

Convergent discretization exists

Pass criteria: Mathematical proof or certificate filed in spec
S4

Certifiability

Audit time

Computable error bounds exist

Pass criteria: Mathematical proof or certificate filed in spec
Imaging

G1 – G3: Imaging Triad

Domain-specific gates for computational imaging modalities. They check whether the inverse problem is fundamentally solvable given the measurement setup.

G1

Recoverability

Imaging

Information-theoretic recovery limit

G2

Carrier Budget

Imaging

Measurement capacity (photon count, k-space)

G3

Operator Mismatch

Imaging

Forward model accuracy

CT QC

CT Quality Control Gates

Application-specific quality control gates for CT imaging. These detect temporal drift, reconstruction artifacts, and threshold breaches relative to clinical limits.

drift

Drift Detection

CT QC

Temporal drift from baseline

artifact

Artifact Detection

CT QC

Ring, cupping, streak artifacts

threshold

Threshold Breach

CT QC

QC metric exceeds clinical limit

How Gates Work

1

Submit

Solver, dataset, or claim is submitted via CLI or web UI

2

Auto-Check

R1-R4 gates run automatically on every submission

3

Domain Check

Imaging triad and QC gates run for applicable modalities

4

Accept / Reject

All gates pass: accepted. Any failure: rejected with report

Propose a New Gate (RFC)

Judge-Rule Authors can propose new gates via the RFC process. Submit below to add your proposal to the review queue.

Log in to submit a gate RFC proposal.