# QPC Joint Structure Challenge — L6 Public Dare Kit

## The dare

Run **separable control only** (K independent jobs, **no transjunctions**).
If you reproduce our **coupled ICC gap ≥ 0.05** without one coupled circuit, QPC withdraws the architecture claim for that instance.

## What we publish

- `l1_instances.json` — fixed seeds, weights, coupling strengths, circuit parameters
- Coupled reference job IDs on IBM Fez (see `results/qpc_jsc_bundle.json`)

## What you must run

1. **Separable arm:** 3 independent circuits per instance (see `separability_core.build_separable`)
2. **Same backend** as our coupled reference (or declare yours)
3. **Same shots** (4096 minimum)
4. Compute ICC with `verify_icc.py` on your counts

## Pass/fail (challenge fails for QPC if you pass)

- Your **separable-only** ICC gap ≥ 0.05 with coupled ICC ≥ 0.02 and ratio ≥ 3×
- Publish all job IDs + raw counts JSON

QPC reference (seed 7, optimized params): coupled raw gap **0.289**; intracontext control gap **0.291** (boundary ICC is depth-independent — high only with cross-context coupling).

## Submit

Email reproducible job IDs to readytogo@quantumpolycontextural.ai

## Verify locally (no IBM needed)

```bash
python3 verify_icc.py --help
```