Majority branch attribution

What was tested?

The increased Majority complexity seen in V36 may be explained by mixing the c5=0 and c5=1 Boolean branches rather than by one stable within-branch mechanism.

Why the test is meaningful

Fixing one Majority input yields different exact functions. When c=0, Maj(a,b,0)=a∧b; when c=1, Maj(a,b,1)=a∨b=a⊕b⊕(a∧b). Stratification can therefore expose branch-mixture confounding.

How it was tested

Partition the exact Majority atlas by c5, attribute affine shell and product terms separately, and compare predefined nonlinear-support summaries in original and fresh cohorts while retaining the pooled result.

What happened

Both branch identities and all composed masks were exact with zero semantic errors over 1,361 original and 300 fresh records. The stratification showed that pooled complexity differences mix branch composition with product/shell behavior; no branch-specific selector was promoted.

Exactness and statistical controls

The Boolean identities are exhaustive. Every stratified cell and empty/small cell was retained in the reports rather than imputed. The fresh cohort repeated the same branch definitions.

What the result means

V37 detects a confounding mechanism in an aggregate contrast. It improves inference quality but does not make either Majority branch cheaper or predict which candidates survive.

Limitations

• Some branch-family cells are small.
• Stratification after observing outcome remains descriptive.
• No online, context-free branch selector or cost advantage was demonstrated.

Evidence trail

Evaluate both exact identities on all lanes, stratify with the frozen c5 definition, report every cell count and repeat the identical decomposition on the fresh cohort.

Canonical variants

Source: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.