Tooling: faster ExclusionCircuit and DisjointTensor support masks with identical receipts

[VictorLiwentao]

Change and scope

At base 0b1cccf86e2f5132ed286a26ef0a681a495740ea, three leave-two-style expected-support genexprs dominate historical verify wall time:

1. ExclusionCircuit.verify — pair leave-two-out over enumerate(self.inputs)
2. ExclusionCircuit.verify_embedding — reverse-frame leave-two over sorted(outputs)
3. DisjointTensor.verify — leave-intersecting inputs over k-set bitmasks (sum(1<<i for i,S if not S&T))

This PR replaces each with an equivalent closed form (full ^ (vertex[a]|vertex[b]) for the ExclusionCircuit paths; full ^ OR_{p in T} vertex[p] for DisjointTensor), keeps every support/topology check, and leaves circuit_sha256 / dag_sha256 / embedding receipts fieldwise identical.

Certificate source_sha256 updates are pin-only:

• scripts/exclusion_circuit.py on the four ExclusionCircuit binders
• scripts/experiments/disjoint_tensor.py on its four binders
• Shared JSON files carry both pins so the rewrites do not conflict

Tooling improvement; no new kappa. Selected frontier untouched.

Result

On Linux / CPython 3.12.3 (Linux-6.12.94+-x86_64-with-glibc2.39), five alternating fresh-process pairs:

Path	Median before → after
ExclusionCircuit(45).verify	0.092 → 0.008 s (11.08×)
ExclusionCircuit(45).verify_embedding	0.323 → 0.059 s (5.46×)
DisjointTensor(26).verify	1.101 → 0.606 s (1.82×)

Timing is host-descriptive and is not a correctness gate.

Packages: research/exclusion-support-masks/, research/exclusion-embedding-masks/, research/disjoint-tensor-support-masks/.

Validation

python3.12 -I -B research/exclusion-support-masks/verify.py
python3.12 -I -O -B research/exclusion-support-masks/verify.py
python3.12 -B scripts/research_package.py --package research/exclusion-support-masks
python3.12 -I -B research/exclusion-embedding-masks/verify.py
python3.12 -I -O -B research/exclusion-embedding-masks/verify.py
python3.12 -B scripts/research_package.py --package research/exclusion-embedding-masks
python3.12 -I -B research/disjoint-tensor-support-masks/verify.py
python3.12 -I -O -B research/disjoint-tensor-support-masks/verify.py
python3.12 -B scripts/research_package.py --package research/disjoint-tensor-support-masks

Expected:

PASS_EXCLUSION_SUPPORT_MASKS …
PASS_EXCLUSION_EMBEDDING_MASKS …
PASS_DISJOINT_TENSOR_SUPPORT_MASKS verify_cases=6 mask_cases=7416 certs=4 no_new_kappa
PASS_REGISTERED_RESEARCH_PACKAGE … outcome=completed

Default verify is read-only. Historical assert/require gates behave as before under -O; digest identity and certificate pin discipline remain the gates. Full make verify, selected cold replay, Lean, and multiplication-theorem obligations were not rerun for these isolated historical-checker changes.

Overlap and credit

Closest priors (different objects): CrocSwap/integer-mult-bounds#756, CrocSwap/integer-mult-bounds#1793, CrocSwap/integer-mult-bounds#1861, CrocSwap/integer-mult-bounds#1922, CrocSwap/integer-mult-bounds#1984, CrocSwap/integer-mult-bounds#1994. DisjointTensor.verify fully overrides the parent and does not call ExclusionCircuit.verify; the closed-form technique is shared, the object and file are not.

Douglas R. Colkitt authored the exclusion / shared-point / disjoint-tensor checkers and bit-compression screens. Prepared by Wentao Li (VictorLiwentao, @Is__Victor) with substantial AI assistance, under Apache-2.0.

[Mister-Meeseeks]

Batch review at cb80ae85dc72b72e7e0e9d8b3a6fc7408e7fa956.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: Support-bitmask and embedding optimizations reproduce prior exclusion/disjoint-tensor verification results. Three independent differential packages and regression suites pass; five source-bound certificate updates change provenance hashes for the modified scripts. Combined compatibility with the digest change is checked separately before integration.

The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions.

Batch review, attribution and evidence. Thank you for the contribution.