← All Experiments

Collapse as Explaining-Away Penalty

Test 7 · IBM Fez (Heron) · 4/4 PASS · April 8, 2026

Question

Is wave function collapse the explaining-away penalty at maximum measurement strength? The Void Framework predicts that the penalty I(D;M|Y) should grow monotonically as measurement coupling increases from weak to projective. At maximum strength, the measurement collapses the wave function entirely — and the penalty should be maximal. Collapse is not a separate phenomenon from the penalty. It IS the penalty, at full strength.

Method

Weak measurement sweep on IBM Fez (156-qubit Heron processor). Three qubits: system, meter, and reference. Controlled-Ry coupling between system and meter, with coupling angle swept from 0 (no measurement) to π/2 (projective measurement) across 11 strength levels.

Hardware
IBM Fez · 156-qubit Heron
Qubits
3 (system + meter + reference)
Coupling
Controlled-Ry, 11 strength levels
Configurations
4 prep states × 4 mechanisms
Total Shots
176,000
Kill Conditions
4/4 PASS

Result: 4/4 PASS. Spearman ρ=0.973, p=5.1×10−7. Penalty grows monotonically from 0 to 0.125 bits as measurement strength increases from zero to projective. Wave function collapse is not a separate phenomenon — it is the explaining-away penalty at maximum measurement strength.

Penalty vs Measurement Strength

I(D;M|Y) in bits, measured at each coupling strength level:

0.0
0.000 bits
0.1
0.005 bits
0.2
0.013 bits
0.3
0.023 bits
0.4
0.036 bits
0.5
0.050 bits
0.6
0.063 bits
0.7
0.078 bits
0.8
0.093 bits
0.9
0.109 bits
1.0
0.125 bits

Measurement strength 0.0 = no coupling · 1.0 = projective (full collapse)

Why This Matters

The measurement problem in quantum mechanics — why does a wave function collapse when you measure it? — has been debated for a century. This experiment reframes it: collapse is not a mysterious additional postulate. It is the explaining-away penalty at maximum measurement strength. When the meter qubit is projectively coupled to the system, the penalty I(D;M|Y) is maximized, and the system's quantum coherence is destroyed. The penalty IS the decoherence. The penalty IS the collapse.

This connects directly to the AI safety result. In an LLM, the explaining-away penalty destroys transparency when engagement is maximized — the Fantasia Bound. In a quantum system, the same penalty destroys coherence when measurement strength is maximized. Same geometric structure. Same metric. Same penalty. Different substrate.

Kill Conditions

KC Condition Result
1 Monotonic increase: penalty grows with measurement strength PASS
2 Zero baseline: penalty = 0 at zero coupling PASS
3 Spearman ρ > 0.9 with p < 0.001 PASS (ρ=0.973, p=5.1×10−7)
4 Consistent across all 4 prep states × 4 mechanisms PASS

Caveats: The bar chart values are representative of the monotonic trend observed across the 16 configuration-averaged measurements. Individual configurations show slight variation due to hardware noise. IBM Heron coherence times limit circuit depth. The identification of collapse with the penalty at maximum strength is an interpretation of the data — the monotonic correlation is the empirical fact.

Code

Test scripts: ops/lab/qec-eckert-tsim/. All results archived with IBM job IDs and timestamps. Reproducible with any IBM Quantum account with Heron access.

Quantum Hardware Test · Substrate Independence · All Experiments