ClaimStab-QC
ClaimStab-QC is a claim-centric framework for checking whether paper conclusions remain valid under software-visible perturbations in quantum software pipelines.
Start Here (2-3 Minutes)
If this is your first visit, follow this order:
Locked run set: - Experiment Matrix
What ClaimStab Checks
ClaimStab evaluates claim outcomes (not only raw scores) under sampled perturbation configurations and returns conservative decisions:
stable: CI lower bound >= thresholdunstable: CI upper bound < thresholdinconclusive: otherwise
Supported claim types:
rankingdecisiondistribution
Canonical schema: - claimstab/spec/schema_v1.json
Why This Matters
Experimental quantum-software conclusions are often sensitive to compiler/sampling settings. ClaimStab makes this explicit by:
- defining claims in executable form,
- evaluating them across perturbation spaces,
- quantifying uncertainty with confidence intervals,
- emitting auditable evidence and reproducible artifacts.
Core Outputs
| File | Purpose |
|---|---|
claim_stability.json |
Per-experiment claim outcomes, CI, decisions, and evidence links |
rq_summary.json |
Aggregated RQ summaries and diagnostics |
robustness_map.json |
Conditional stability cells + robust core/frontier/lockdown summaries |
scores.csv |
Raw score/evaluation rows with timing metadata |
stability_report.html |
Human-readable report |
Snapshot
Current empirical signal
compilation_onlytends to be the most stable space.sampling_onlyis the strongest instability driver.combined_lightstill exposes near-tie fragility.
Explore / Community Capabilities (Advanced)
These are preserved as future-facing infrastructure assets, but are not required for first-time onboarding: