EXP-005EXPERIMENTAL
Advantage Decomposition / Capability portfolio

Hybrid semantic portfolio

A portfolio of exact specialized representations can cover more contexts than one global grammar.

Claim statusNO MINING ADVANTAGE CLAIM

Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.

Date2026-08-14
HardwareClassical deterministic verifier
Run scopeSeven entries including repaired rerun
ReproducibilityIncludes original V5A, corrected V5A-R1 and independent validation variants.
01 / Question & hypothesis

What was tested?

A portfolio of exact specialized representations can cover more contexts than one global grammar.

02 / Scientific basis

Why the test is meaningful

A capability portfolio is a union of separately scoped exact representations. Scientific validity requires that a dispatcher never apply a capability outside its proven domain and that failed candidates remain in the audit trail. V5 therefore treats invalidation as data rather than silently optimizing it away.

P = {(Cⱼ,Dⱼ)} where Cⱼ is the certified domaindispatch(x)=j only if x∈Cⱼsoundness ⇔ ∀x∈Cⱼ: Dⱼ(x)=D_baseline(x)
Hybrid semantic portfolioVisual reading of the published metrics and gates for EXP-005; it summarizes the registered result, not a mining advantage.EXP-005 / HYBRID SEMANTIC PORTFOLIOBOUNDED INVALIDATIONS1CORRECTED RERUNV5A-R1CLAIMNONE
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-005; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Test candidate capability types, invalidate the original V5A where its bounded premise failed, rerun the corrected V5A-R1, and compile only exact survivors.

04 / Observed result

What happened

Bounded invalidations1
Corrected rerunV5A-R1
ClaimNONE

The family completed after a bounded invalidation and repaired rerun. The invalid result is retained in the provenance rather than erased.

05 / Validation

Exactness and statistical controls

The original V5A failure was bounded, recorded and replaced by the corrected V5A-R1; independent variants then checked only surviving capabilities. This guards against survivorship bias and post-hoc deletion of a negative branch.

06 / Interpretation

What the result means

The experiment supports disciplined capability composition and demonstrates why invalidated exploratory branches must remain visible.

Limitations

  • Portfolio coverage may remain sparse.
  • Dispatcher and certificate checks count toward runtime.
  • Repairing one invalid premise does not validate untested capability families.
07 / Reproduction

Evidence trail

Includes original V5A, corrected V5A-R1 and independent validation variants.

Canonical variants

SUBENGINE-V5ASUBENGINE-V5A-R1SUBENGINE-V5BSUBENGINE-V5CSUBENGINE-V5DSUBENGINE-V5ESUBENGINE-V5F

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