Skip to content

QSA factorized phase coherence and bounded Pauli readout - #85

Draft
Inserian wants to merge 4 commits into
agent/qsa-exact-linear-language-v1from
agent/qsa-phase-coherence-v1
Draft

QSA factorized phase coherence and bounded Pauli readout#85
Inserian wants to merge 4 commits into
agent/qsa-exact-linear-language-v1from
agent/qsa-phase-coherence-v1

Conversation

@Inserian

@Inserian Inserian commented Aug 15, 2026

Copy link
Copy Markdown
Member

Adds the first QSA-native coherent readout primitive intended for Brain Cortex VNext without moving semantic authority into QSA.

The branch is a sibling of accepted PR83, not a child of the active polynomial lane. It extends PhaseGraphState so compact uniform quadratic phase states can expose physically meaningful interference/correlation observables without materializing the 2^n statevector:

  • O(n+E) equatorial coherence for every logical qubit, with Re(c_i)=<X_i> and Im(c_i)=<Y_i>;
  • bounded multi-body Pauli X/Y/Z expectation using exact structural factorization over variables outside the flipped support and enumeration only over the configured flipped support;
  • explicit receipts proving factorized execution, edge/support accounting, assignment count, and zero dense materialization;
  • fail-closed duplicate-support, flip-support and enumeration limits;
  • randomized differential validation against dense QRegister + PauliObservable controls;
  • explicit three-body X-X-X Tripair correlation;
  • continued 100,000-logical-qubit / 99,999-edge structural coverage, now including coherence and local Tripair readout.

Exact-head acceptance evidence

Accepted research head: f0eba309f2519019e8ac524239fc7b5dd69dccdd

  • Structural Hardening 31905522965: success
  • Build and Test 31905522963: success
  • Exact Low-H Phase Graph Sum 31905523013: success
  • QSA 0.2.0 Runtime Evidence 31905523089: success
  • sanitizer jobs: success
  • randomized native differential validation: success
  • Windows/Linux/macOS build/test jobs: success

The 320 randomized 2-9-qubit phase-graph sequences retain full-state comparison to QRegister; new equatorial X/Y values and randomized X/Y/Z products up to four factors are independently compared with dense PauliObservable. The 100,000-logical-qubit test remains below the existing 20 MiB structural gate, and a local Tripair XXX query uses eight flipped-support assignments rather than global state enumeration.

This closes a specific gap in the existing phase graph: it already retained coherent phase structure compactly, but the readout surface was primarily computational-basis probability/amplitude/materialization. Brain cannot exploit route interference if coherent route state is discarded before recombination. This branch exposes that recombination information directly while keeping Brain in charge of semantic meaning and final use.

No raw language enters QSA. No Brain source, training, protected state, production routing, or semantic authority is changed. The frozen Main Brain integration contract requires a separate pinned install and disabled -> shadow -> exact-zero trainable adapter; the current V10 QSA bundle is not overwritten.

Research-accepted on this exact head, still draft/unmerged. This is a validated structured quantum-state primitive on classical hardware, not a claim of universal quantum advantage or Brain benchmark improvement before the Main Brain integration decision.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant