Skip to content

GOTM-SCPN Paper 0 First Preregistered Downstream Experiment

Status: preregistered no-QPU measured-system design, dated 2026-05-19.

This document selects the first downstream experiment from the GOTM-SCPN Paper 0: The Foundational Framework pathway. It is a preregistration contract, not a result report. It does not claim physical validation, biological validation, quantum advantage, or hardware readiness.

Selected lane

Field Preregistered value
Lane family K_nm causal-efficacy and coupling-affinity
Evidence class before execution measured-system candidate
Primary system EEG alpha PLV eight-channel coupling candidate
Cross-domain negative control IEEE 5-bus power-grid sparse coupling candidate
Source programme GOTM-SCPN Paper 0: The Foundational Framework downstream pathway
Registry evidence docs/paper0_lane_registry.md; data/paper0_lane_registry.json
Existing gap contract docs/internal/audits/contracts/knm_physical_validation_contract_2026-04-30.md
Hardware status no QPU submission; no hardware execution; no IBM spend

Scientific question

Can the repository's K_nm coupling construction produce a reproducible, unit-aware, uncertainty-aware measured-system replay that distinguishes a plausible coupling-affinity candidate from a sparse negative-control topology without promoting topology correlation into physical magnitude validation?

Hypotheses

ID Statement Promotion boundary
H1 The EEG alpha PLV candidate will retain positive rank-order association with source-derived K_nm coupling after the preregistered normalisation and uncertainty model are applied. Rank-order association is not physical magnitude validation.
H2 Direct magnitude error will remain non-closing unless scale, units, and uncertainty pass the acceptance gates below. Magnitude validation remains blocked if any gate fails.
H3 The IEEE 5-bus negative-control candidate will not satisfy the same acceptance gates under the same replay procedure. Failure of the negative control only supports specificity of the replay protocol, not broad validation.

Inputs

Required committed inputs:

  • docs/internal/audits/contracts/knm_physical_validation_contract_2026-04-30.md
  • data/public_application_benchmarks/
  • data/knm_physical_validation/measured_couplings_power_grid_ieee5bus.json
  • docs/paper0_experimental_pathway.md
  • docs/paper0_lane_registry.md

The EEG candidate currently appears in the K_nm physical-validation contract as eeg_alpha_plv_8ch. The power-grid candidate appears as ieee5bus_power_grid. Both are pre-existing candidates and remain non-closing until this preregistered replay is implemented and evaluated.

Locked analysis plan

  1. Load the K_nm candidate audit rows from the existing physical-validation artefacts.
  2. Build a measured-edge table for the EEG alpha PLV candidate with explicit units, normalisation, and uncertainty fields.
  3. Re-run the same table schema for the IEEE 5-bus negative-control candidate.
  4. Compute rank-order association, direct magnitude error, and best scale-through-origin error for both candidates.
  5. Run null models that preserve node count and edge-density class.
  6. Emit a machine-readable replay artefact and a Markdown report.
  7. Keep the K_nm physical-validation gap open unless every acceptance gate passes.

Acceptance gates

Gate Pass condition Failure consequence
Named system Candidate row names a concrete measured system and data provenance. Block measured-system promotion.
Units Every measured edge records units or a documented dimensionless convention. Block magnitude claims.
Normalisation Normalisation from measured coupling to K_nm comparison space is explicit and reproducible. Block magnitude claims.
Uncertainty Every measured edge records uncertainty or a conservative uncertainty model. Block magnitude claims.
Pairwise coverage Required pairwise edges are either measured or explicitly censored with a rule. Block full-system claims.
Null models Null replay rows are generated under fixed random seeds and topology constraints. Block specificity claims.
Negative control IEEE 5-bus candidate is evaluated under the same procedure. Block cross-domain specificity claims.
Reproducibility JSON and Markdown replay artefacts are regenerated by a command. Block publication use.

Falsifiers

The experiment fails to promote beyond measured-system candidate if any of the following occurs:

  • the EEG candidate lacks explicit units or a defensible dimensionless convention;
  • uncertainty is missing for measured edges;
  • the negative-control candidate passes the same acceptance gates with equal or better evidence;
  • null models match or exceed the candidate association metrics;
  • direct magnitude error remains unconstrained after scale and uncertainty are applied;
  • any replay artefact cannot be regenerated deterministically.

Claim boundary

A passing replay may support only this bounded statement:

The preregistered K_nm measured-system replay is reproducible for the named candidate under the declared units, uncertainty, null-model, and negative- control gates.

It may not claim that GOTM-SCPN Paper 0: The Foundational Framework is externally validated. It may not claim broad K_nm physical validation, quantum advantage, biological truth, or hardware/QPU readiness.

Required implementation artefacts

The next implementation slice should add:

Artefact Purpose
scripts/run_paper0_knm_preregistered_replay.py Deterministic no-QPU replay command.
data/paper0_knm_preregistered_replay.json Machine-readable replay result.
docs/paper0_knm_preregistered_replay.md Human-readable replay report.
tests/test_paper0_knm_preregistered_replay.py Acceptance-gate and falsifier tests.
scpn-bench paper0-knm-preregistered-replay-gate Release-safe reproducibility gate.

Hardware boundary

No QPU execution is part of this preregistration. Additional IBM or hardware shots cannot close this measured-system coupling-magnitude gap unless the claim is explicitly changed to a backend-device coupling-map claim with calibration provenance.

Implemented replay gate

The preregistered design is now bound to a deterministic no-QPU replay gate:

  • scripts/run_paper0_knm_preregistered_replay.py
  • scripts/compare_paper0_knm_preregistered_replay.py
  • scripts/run_paper0_knm_preregistered_replay_gate.py
  • data/paper0_knm_preregistered_replay.json
  • docs/paper0_knm_preregistered_replay.md
  • tests/test_paper0_knm_preregistered_replay.py
  • scpn-bench paper0-knm-preregistered-replay-gate

The replay is intentionally non-closing. It preserves the primary EEG alpha PLV uncertainty blocker, exercises the IEEE 5-bus negative control, adds a deterministic permutation-null battery for matrix-alignment diagnostics, records input SHA-256 digests for replay provenance, emits an explicit do-not-promote decision object, and keeps QPU submission blocked until measured coupling magnitudes with uncertainty are available.