Skip to content

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:

  1. Quickstart
  2. Reproduce
  3. Output Directory Map

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 >= threshold
  • unstable: CI upper bound < threshold
  • inconclusive: otherwise

Supported claim types:

  • ranking
  • decision
  • distribution

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:

  1. defining claims in executable form,
  2. evaluating them across perturbation spaces,
  3. quantifying uncertainty with confidence intervals,
  4. 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_only tends to be the most stable space.
  • sampling_only is the strongest instability driver.
  • combined_light still exposes near-tie fragility.

ClaimStab Pipeline

Explore / Community Capabilities (Advanced)

These are preserved as future-facing infrastructure assets, but are not required for first-time onboarding: