Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
311 commits
Select commit Hold shift + click to select a range
cc2b789
fix: use exact Klein numeric conversion
ZiboJin Jul 28, 2026
678e4dc
test: specify shared Klein numeric matrix
ZiboJin Jul 28, 2026
2d29f3a
perf: share Klein numeric matrix across tasks
ZiboJin Jul 28, 2026
8ba8d91
fix: harden shared Klein matrix initialization
ZiboJin Jul 28, 2026
24c80c4
docs: harden R01 experiment evidence gates
ZiboJin Jul 28, 2026
a9cf539
test: specify R01 compact fixture replay
ZiboJin Jul 28, 2026
30e60c8
research: add exact R01 number-conserving fixture
ZiboJin Jul 28, 2026
d4b8077
research: record six-mode number-conserving Klein result
ZiboJin Jul 28, 2026
d42786a
docs: harden R01 BdG evidence protocol
ZiboJin Jul 28, 2026
fe301b5
test: specify R01 fixture schema v2
ZiboJin Jul 28, 2026
6686cb2
research: add exact R01 BdG fixture
ZiboJin Jul 28, 2026
8ad4553
docs: record exact R01 BdG no-go
ZiboJin Jul 28, 2026
a48c525
docs: record distributed R01 replay
ZiboJin Jul 28, 2026
1fc9342
test: require raw-backed R01 evidence gate
ZiboJin Jul 28, 2026
b049bed
test: close R01 evidence provenance gaps
ZiboJin Jul 28, 2026
dd44e09
test: bind R01 fixture to raw evidence
ZiboJin Jul 28, 2026
408266c
fix: validate R01 fixture against raw evidence
ZiboJin Jul 28, 2026
6a608c1
docs: amend Task 9 for the exact no-go branch
ZiboJin Jul 29, 2026
f5aa9df
test: specify exact R01 no-go classification
ZiboJin Jul 29, 2026
00d8ca8
docs: design exterior cone discovery loop
ZiboJin Jul 29, 2026
92e3677
docs: plan exterior cone throughput loop
ZiboJin Jul 29, 2026
4eaceb5
test: specify exterior cone certificates
ZiboJin Jul 29, 2026
01f98dc
feat: add exterior cone certificate search
ZiboJin Jul 29, 2026
d5b5fa6
test: harden exterior cone certificate contract
ZiboJin Jul 29, 2026
86cbb90
docs: amend exterior loop for exact-first search
ZiboJin Jul 29, 2026
ea616ea
fix: keep exterior cone certificates replayable
ZiboJin Jul 29, 2026
16f3f36
test: specify exact exterior candidate cards
ZiboJin Jul 29, 2026
1d788b4
docs: pivot sign-free search to overlap-first cones
xianzhipan Jul 29, 2026
c3ed9d1
docs: record exterior oracle review loop
ZiboJin Jul 29, 2026
0f32d93
feat: add exact occupation-basis quadratic algebra
ZiboJin Jul 28, 2026
7e24bad
feat: encode exact Klein-Hodge Fock transform
ZiboJin Jul 28, 2026
3cfa6ee
feat: add exact exterior candidate grammars
ZiboJin Jul 29, 2026
c73b8c9
test: specify thin exterior first scan
ZiboJin Jul 29, 2026
c7f373b
test: align injected atoms with candidate dimension
ZiboJin Jul 29, 2026
d0e2950
feat: add thin exterior survivor scan
ZiboJin Jul 29, 2026
f46f7d7
test: require unique canonical exterior candidates
ZiboJin Jul 29, 2026
5f92b1d
research: add six-mode Fock CP screen
xianzhipan Jul 29, 2026
0168adc
fix: make exterior candidate payloads canonical
ZiboJin Jul 29, 2026
cb04e61
research: close depth-two Klein Fock CP screen
xianzhipan Jul 29, 2026
bba01fb
test: require launch-safe exterior scan
ZiboJin Jul 29, 2026
37f69a3
research: build tensor-square plaquette oracle
xianzhipan Jul 29, 2026
b90a506
fix: make exterior scan launch safe
ZiboJin Jul 29, 2026
b5945e3
research: close tensor-square physical mapping
xianzhipan Jul 29, 2026
3ee4451
plan: freeze survivor pressure stage
ZiboJin Jul 29, 2026
b4d50af
research: add edge-gauge cocycle oracle
xianzhipan Jul 29, 2026
bba0f79
research: close affine edge-gauge cocycle
xianzhipan Jul 29, 2026
eda8002
test: specify survivor pressure protocol
ZiboJin Jul 29, 2026
d5a709e
feat: add survivor pressure protocol
ZiboJin Jul 29, 2026
a54df88
test: correct pressure fixtures
ZiboJin Jul 29, 2026
f588ce4
test: lock survivor pressure lineage
ZiboJin Jul 29, 2026
36c79f5
fix: enforce survivor pressure lineage
ZiboJin Jul 29, 2026
e4572c3
fix: bind smoke specs to smoke protocol
ZiboJin Jul 29, 2026
9bb9806
test: lock reproduced parent hash evidence
ZiboJin Jul 29, 2026
c4919c4
fix: validate reproduced survivor lineage
ZiboJin Jul 29, 2026
fe3b155
docs: record survivor pressure production
ZiboJin Jul 29, 2026
216e1bb
feat: replay pressure uncertainties at high precision
ZiboJin Jul 29, 2026
30f6df9
feat: add depth-12 survivor scan
ZiboJin Jul 29, 2026
dcdf26b
docs: record exact replay and depth-12 scan
ZiboJin Jul 29, 2026
859347d
feat: continue exact-fallback survivors to depth eight
ZiboJin Jul 29, 2026
42a5e53
docs: open unconventional model discovery
xianzhipan Jul 29, 2026
070cca4
feat: replay depth-twelve uncertainties exactly
ZiboJin Jul 29, 2026
77acaf2
feat: continue exterior survivors to depth 16
ZiboJin Jul 29, 2026
8f4a388
feat: pressure-test exact depth-eight survivors
ZiboJin Jul 29, 2026
bb99ca5
feat: complete high-precision depth-twelve tranche
ZiboJin Jul 29, 2026
5de0ee5
feat(qmc): replay depth16 uncertain failures
ZiboJin Jul 29, 2026
e02bf96
docs(qmc): record deep exterior pressure results
ZiboJin Jul 29, 2026
59c9459
test(qmc): specify depth16 structural ranking
ZiboJin Jul 29, 2026
78bc511
feat(qmc): invert exterior cards to Hermitian HS models
ZiboJin Jul 29, 2026
427c52d
feat(qmc): rank depth16 survivors structurally
ZiboJin Jul 29, 2026
8ac2cfe
feat(qmc): search exact5 shared exterior cones
ZiboJin Jul 29, 2026
da7a2ec
docs(qmc): record structural and inverse-H advances
ZiboJin Jul 29, 2026
74d9c82
feat(qmc): continue depth16 HP survivors exactly
ZiboJin Jul 29, 2026
e786783
feat(qmc): search redundant exact5 invariant cones
ZiboJin Jul 29, 2026
f47ef2e
fix(qmc): normalize tiny cone rays before replay
ZiboJin Jul 29, 2026
d272e81
docs(qmc): record exact continuations and cone audit
ZiboJin Jul 29, 2026
c375720
audit(qmc): scope seed61 redundant cone certificate
ZiboJin Jul 29, 2026
d8dc6f4
test(qmc): specify continuation structural ranking
ZiboJin Jul 29, 2026
a06a1c5
feat(qmc): rank depth12 continuations structurally
ZiboJin Jul 29, 2026
567f0b8
docs(qmc): record exact depth12 continuations
ZiboJin Jul 29, 2026
40ad5c4
feat(qmc): audit seed61 positive realization
ZiboJin Jul 29, 2026
82bdf48
test(qmc): accept earliest cancellation witness
ZiboJin Jul 29, 2026
3f018b2
feat(qmc): search trace-compatible full-Fock cones
ZiboJin Jul 29, 2026
35ada6f
docs(qmc): record structural early stops
ZiboJin Jul 29, 2026
71a8593
docs(qmc): record structural early stops
ZiboJin Jul 29, 2026
860603a
docs: audit affine A4 Fock positivity route
ZiboJin Jul 29, 2026
bee8a1d
perf(qmc): sum exact Fock grade traces
ZiboJin Jul 29, 2026
72460db
perf(qmc): defer exact full-Fock rank gate
ZiboJin Jul 29, 2026
0fe5020
fix(qmc): serialize full-Fock coupling count
ZiboJin Jul 29, 2026
83e7b36
chore(qmc): report full-Fock search terminal reason
ZiboJin Jul 29, 2026
8be4a36
docs(qmc): record full-Fock route triage
ZiboJin Jul 29, 2026
106fc7e
merge: retain published research history
ZiboJin Jul 29, 2026
d9bdb74
research: build unconventional model factory
xianzhipan Jul 29, 2026
083751f
docs(qmc): audit seed61 Hodge spinor cone
ZiboJin Jul 29, 2026
51d733e
docs(qmc): record Hodge and oddcycle triage
ZiboJin Jul 29, 2026
0eac2ad
feat(qmc): generalize exact-gated cone search
ZiboJin Jul 29, 2026
8033fc8
docs(qmc): stop deep oddcycle simplicial restarts
ZiboJin Jul 29, 2026
c915a68
feat: certify seed61 stable spectral tail
ZiboJin Jul 29, 2026
9c86bfa
docs(qmc): retire oddcycle sector cones
ZiboJin Jul 29, 2026
26b0d98
research: resolve three unconventional candidates
xianzhipan Jul 29, 2026
bf2aeef
prove seed61 top-pair product tail
ZiboJin Jul 29, 2026
c98b913
fix(qmc): freeze exact seed61 counterexample
ZiboJin Jul 29, 2026
38e04a5
feat(qmc): verify seed61 short words exactly
ZiboJin Jul 29, 2026
b660bb7
docs(qmc): retire seed61 by exact counterexample
ZiboJin Jul 29, 2026
ffa609d
feat(qmc): target exact short-word scans
ZiboJin Jul 29, 2026
4daabad
feat(qmc): add exact-gated longword adversary
ZiboJin Jul 29, 2026
c9c138c
docs(qmc): finalize depth16 HP continuation
ZiboJin Jul 29, 2026
3de7b61
docs(qmc): certify oddcycle seeds through depth23
ZiboJin Jul 29, 2026
a0ab66e
docs(qmc): record longword oddcycle survivors
ZiboJin Jul 29, 2026
2ae03be
fix(qmc): rerank longword adversaries at high precision
ZiboJin Jul 29, 2026
f251ddc
docs: add complete project master summary
xianzhipan Jul 29, 2026
10bfc6a
fix(qmc): serialize unbounded exact weights
ZiboJin Jul 29, 2026
f4512c9
Merge remote-tracking branch 'shared/research/no-negative-vibes' into…
xianzhipan Jul 29, 2026
7da964d
docs(qmc): record exact depth27 continuation
ZiboJin Jul 29, 2026
349d8b3
docs: integrate collaborator research status
xianzhipan Jul 29, 2026
34f4da1
docs: refresh collaborator branch audit
xianzhipan Jul 29, 2026
2f0f20d
docs(qmc): validate high precision longword grid
ZiboJin Jul 29, 2026
b4e2185
feat(qmc): freeze symmetric oddcycle cone lemmas
ZiboJin Jul 29, 2026
ae95737
docs(qmc): record symmetric oddcycle cone branch
ZiboJin Jul 29, 2026
99b09b9
docs(qmc): close deep longword campaign
ZiboJin Jul 29, 2026
23bbe4a
feat(qmc): scan fixed symmetric oddcycle exactly
ZiboJin Jul 29, 2026
f4188a7
test(qmc): freeze oddcycle chamber counterexample
ZiboJin Jul 29, 2026
841ff6b
test(qmc): audit symmetric oddcycle complement
ZiboJin Jul 29, 2026
9583e91
test(qmc): freeze fixed invariant chamber obstruction
ZiboJin Jul 29, 2026
dd11495
test(qmc): certify unit-winding Bernstein margins
ZiboJin Jul 29, 2026
8f5a823
test(qmc): freeze endpoint semigroup counterexample
ZiboJin Jul 29, 2026
ca40bb6
test(qmc): freeze odd-sector split obstruction
ZiboJin Jul 29, 2026
0538421
test(qmc): stop invalid complement cone splits
ZiboJin Jul 29, 2026
192142f
docs(qmc): record fixed depth27 and split obstructions
ZiboJin Jul 29, 2026
9ca0c3d
feat(qmc): search fixed full Fock cone exactly
ZiboJin Jul 29, 2026
0aa0e3c
feat(qmc): certify pure and reflection-square words
ZiboJin Jul 29, 2026
14e8c81
docs(qmc): record infinite subclasses and full Fock search
ZiboJin Jul 29, 2026
974cab6
feat(qmc): precondition fixed full Fock cone
ZiboJin Jul 29, 2026
cd4d57c
feat(qmc): warm-start grade24 full Fock cone
ZiboJin Jul 29, 2026
84d757e
feat(qmc): certify fixed oddcycle words at all lengths
ZiboJin Jul 29, 2026
9f1a13a
ts-phase:_validate_determinant_identity
ZiboJin Jul 29, 2026
ed8407c
docs(qmc): record arbitrary-word theorem
ZiboJin Jul 29, 2026
b482419
docs(qmc): hash warm full Fock archive
ZiboJin Jul 29, 2026
08a9c42
ts-phase:add-m3-m4-ED-pilot
ZiboJin Jul 29, 2026
a39f59d
feat(qmc): certify oddcycle interval alphabet
ZiboJin Jul 29, 2026
2a1348b
docs(qmc): record continuum alphabet theorem
ZiboJin Jul 29, 2026
bb44271
docs(qmc): audit oddcycle novelty reductions
ZiboJin Jul 29, 2026
b54837c
feat(qmc): realize oddcycle theorem as physical transfer
ZiboJin Jul 29, 2026
ef73bb6
docs(qmc): record physical transfer realization
ZiboJin Jul 29, 2026
f561d9a
docs(qmc): audit oddcycle challenge completion
ZiboJin Jul 29, 2026
4a96fc8
feat(qmc): add fast oddcycle discovery screen
ZiboJin Jul 29, 2026
efb2e18
ts-phase:add-stabilized-DQMC-validation
ZiboJin Jul 29, 2026
264650e
feat(qmc): batch oddcycle discovery domains
ZiboJin Jul 29, 2026
5986da2
feat(qmc): add full oddcycle metric filter
ZiboJin Jul 29, 2026
05317cc
feat(qmc): scan oddcycle joint-metric frontier
ZiboJin Jul 29, 2026
5e8cefb
ts-phase:add-dual-machine-coarse-map
ZiboJin Jul 29, 2026
01b3bcd
feat(qmc): bound multi-point oddcycle exteriors
ZiboJin Jul 29, 2026
00cf304
feat(qmc): stress joint oddcycle words exactly
ZiboJin Jul 29, 2026
37ae584
ts-phase:harden-scan-provenance-audits
ZiboJin Jul 29, 2026
c162487
feat(qmc): certify exterior block contractions
ZiboJin Jul 29, 2026
b4459ae
ts-phase:tune-density-audit-tolerance
ZiboJin Jul 29, 2026
838f428
feat(qmc): profile coupled oddcycle tails
ZiboJin Jul 29, 2026
2e46b67
feat(qmc): certify oddcycle common-R contraction
ZiboJin Jul 29, 2026
de6c01e
docs: preregister next positivity search
xianzhipan Jul 29, 2026
597e91b
docs: tighten typed positivity specification
xianzhipan Jul 29, 2026
95af6e0
ts-phase:finalize-clean-coarse-map
ZiboJin Jul 29, 2026
d28931c
ts-phase:fix-coarse-figure-path
ZiboJin Jul 29, 2026
9b91d19
feat(qmc): exclude a common metric exactly
ZiboJin Jul 29, 2026
120d907
feat(qmc): realize the oddcycle pair physically
ZiboJin Jul 29, 2026
3322167
ts-phase:correct-ridge-significance-range
ZiboJin Jul 29, 2026
db97a23
feat(qmc): reduce the oddcycle coupled tail exactly
ZiboJin Jul 29, 2026
cf2867c
feat(qmc): certify finite-state oddcycle positivity
ZiboJin Jul 29, 2026
26c03ba
feat(qmc): close oddcycle novelty and physical gates
ZiboJin Jul 29, 2026
4b7a0bb
feat(qmc): preregister adaptive stage4 scan
ZiboJin Jul 29, 2026
d5646cd
docs(qmc): package the oddcycle publication result
ZiboJin Jul 29, 2026
8037e4a
fix(qmc): enforce stage4 production audits
ZiboJin Jul 29, 2026
bd0bcb0
fix(qmc): authenticate stage4 budget release
ZiboJin Jul 29, 2026
6cf0bf2
docs(qmc): delimit oddcycle prior-art boundary
ZiboJin Jul 29, 2026
84fc08b
docs(qmc): record stage4 production gate
ZiboJin Jul 29, 2026
38e1888
docs(qmc): record oddcycle publication replay lessons
ZiboJin Jul 29, 2026
b08ea2a
feat(qmc): add resumable oddcycle pair frontier scan
ZiboJin Jul 29, 2026
0fbfee1
fix(qmc): harden oddcycle pair scan gates
ZiboJin Jul 29, 2026
3a3fce2
feat(qmc): audit stage4 production ranking
ZiboJin Jul 29, 2026
dc87c6c
feat(qmc): release audited m10 sentinel
ZiboJin Jul 29, 2026
d771273
feat(qmc): add non-induced exterior frontier scan
ZiboJin Jul 29, 2026
8969305
fix(qmc): harden exterior scan replay and stress gates
ZiboJin Jul 29, 2026
b1c2b88
results(qmc): record stage4 statistical stop
ZiboJin Jul 29, 2026
b5cdfca
docs(qmc): record stage4 branch push
ZiboJin Jul 29, 2026
f98b01c
docs(qmc): record remote frontier scan results
ZiboJin Jul 29, 2026
d2026fd
fix(qmc): make frontier scan retries identity-safe
ZiboJin Jul 29, 2026
852a83a
feat(qmc): audit temporal block sampler
ZiboJin Jul 29, 2026
7539e24
results(qmc): stop temporal block sampler
ZiboJin Jul 29, 2026
823aedb
docs(qmc): record exact dual frontier promotions
ZiboJin Jul 29, 2026
db0d114
feat(qmc): validate channel reflection sampler
ZiboJin Jul 29, 2026
29a54f3
docs(qmc): promote robust exact oddcycle candidate
ZiboJin Jul 29, 2026
ebd0883
results(qmc): stop channel reflection sampler
ZiboJin Jul 29, 2026
5330d6a
docs(qmc): checkpoint remote search round
ZiboJin Jul 29, 2026
2472faa
docs(qmc): close channel reflection round
ZiboJin Jul 29, 2026
3ee7905
docs(qmc): record closeout delivery
ZiboJin Jul 29, 2026
0527038
feat(qmc): freeze robust oddcycle certificate
ZiboJin Jul 30, 2026
e3aafd8
docs(qmc): checkpoint robust certificate packaging
ZiboJin Jul 30, 2026
3ac8509
docs(qmc): design full Majorana Wei audit
ZiboJin Jul 30, 2026
4f60575
docs(qmc): fix Nambu lift convention
ZiboJin Jul 30, 2026
f2b7fc2
docs(qmc): checkpoint Majorana Wei audit
ZiboJin Jul 30, 2026
bf43005
feat(qmc): certify oddcycle outside Wei contraction class
ZiboJin Jul 30, 2026
b875afd
feat(qmc): add Majorana Wei publication gate
ZiboJin Jul 30, 2026
2c94654
docs(qmc): prove Majorana Wei novelty boundary
ZiboJin Jul 30, 2026
6aba2a6
fix(qmc): tolerate cold Git provenance replay
ZiboJin Jul 30, 2026
bf8f326
docs(qmc): archive Majorana Wei exact replay
ZiboJin Jul 30, 2026
d69a0d3
docs(qmc): record Majorana Wei remote checkpoint
ZiboJin Jul 30, 2026
32e1b49
docs(qmc): design same-alphabet local H search
ZiboJin Jul 30, 2026
35c7384
docs(qmc): plan local HS first batch
ZiboJin Jul 30, 2026
bf441e2
test(qmc): specify oddcycle word operator compiler
ZiboJin Jul 30, 2026
58961af
feat(qmc): compile oddcycle word operators exactly
ZiboJin Jul 30, 2026
0088fa5
test(qmc): specify same-alphabet word dictionary
ZiboJin Jul 30, 2026
985a55c
feat(qmc): enumerate same-alphabet word-pair rays
ZiboJin Jul 30, 2026
19b7b6c
test(qmc): specify oddcycle local-H cone scan
ZiboJin Jul 30, 2026
079f680
feat(qmc): scan positive oddcycle local-H cones
ZiboJin Jul 30, 2026
c2f9816
test(qmc): specify exact local-H promotion
ZiboJin Jul 30, 2026
4be7a76
feat(qmc): promote local-H cone survivors exactly
ZiboJin Jul 30, 2026
7a4ab59
fix(qmc): distinguish exact local-H promotion outcomes
ZiboJin Jul 30, 2026
aa6a5e6
test(qmc): specify named local H target scan
ZiboJin Jul 30, 2026
e5322e6
feat(qmc): scan named local H targets in word cone
ZiboJin Jul 30, 2026
eefeb1f
test(qmc): specify transfer Hamiltonian portfolio
ZiboJin Jul 30, 2026
cc82c85
feat(qmc): rank same-alphabet transfer Hamiltonians
ZiboJin Jul 30, 2026
d0e67d8
test(qmc): specify honest inconclusive portfolio records
ZiboJin Jul 30, 2026
70a431f
fix(qmc): preserve honest inconclusive portfolio records
ZiboJin Jul 30, 2026
34fe44e
test(qmc): specify resumable local-H batch runner
ZiboJin Jul 30, 2026
7d792fc
feat(qmc): freeze resumable local-H first batch
ZiboJin Jul 30, 2026
73cf75c
test(qmc): specify durable runner failure semantics
ZiboJin Jul 30, 2026
bebab46
fix(qmc): harden resumable runner failure semantics
ZiboJin Jul 30, 2026
f7da2e3
docs(qmc): checkpoint oddcycle local-H first batch
ZiboJin Jul 30, 2026
a6a0110
docs(qmc): freeze Survivor A Hamiltonian search
ZiboJin Jul 30, 2026
511a25b
docs(qmc): freeze Survivor A search plan
ZiboJin Jul 30, 2026
e16b49e
docs(qmc): start positive Hamiltonian reappraisal
xianzhipan Jul 30, 2026
cd89325
research(qmc): close odd block TN via stoquastic gauge
xianzhipan Jul 30, 2026
95fb640
feat(qmc): reconstruct frozen Survivor A exactly
ZiboJin Jul 30, 2026
03aca59
research(qmc): resolve revived monomial and semigroup routes
xianzhipan Jul 30, 2026
06bd75c
research(qmc): revive local orthogonal contraction model
xianzhipan Jul 30, 2026
6871920
research(qmc): exclude simple solvers for orthogonal model
xianzhipan Jul 30, 2026
b3943da
research-qmc-audit-orthogonal-model-Majorana-square
xianzhipan Jul 30, 2026
2f120b0
docs(qmc): checkpoint Hamiltonian portfolio launch
ZiboJin Jul 30, 2026
a493ebf
docs(qmc): record portfolio resume checkpoint
ZiboJin Jul 30, 2026
7c395bf
test(qmc): specify Survivor A precision audit
ZiboJin Jul 30, 2026
37fc3ac
docs(qmc): snapshot running Hamiltonian portfolios
ZiboJin Jul 30, 2026
15a4aab
Merge remote-tracking branch 'teammate/codex/positive-hamiltonian-rea…
ZiboJin Jul 30, 2026
57bc6c7
research-qmc-promote-orthogonal-QNC-candidate
xianzhipan Jul 30, 2026
4531a86
Merge remote-tracking branch 'pr-fork/tensor-square-phase-diagram' in…
ZiboJin Jul 30, 2026
a01ef99
docs(qmc): package complete challenge submission
ZiboJin Jul 30, 2026
f4781bf
Merge remote-tracking branch 'shared/codex/positive-hamiltonian-reapp…
ZiboJin Jul 30, 2026
21b8538
research-qmc-promote-orthogonal-QNC-candidate
xianzhipan Jul 30, 2026
020f894
docs(qmc): report orthogonal QNC promotion
ZiboJin Jul 30, 2026
4d0b96d
docs(qmc): align submission ledger status
ZiboJin Jul 30, 2026
35f0a10
docs(qmc): enrich final challenge submission
xianzhipan Jul 30, 2026
bf46110
docs(qmc): distinguish final oddcycle result
ZiboJin Jul 30, 2026
28aa5a3
merge: reconcile final organizer submission
xianzhipan Jul 30, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
Original file line number Diff line number Diff line change
@@ -0,0 +1,324 @@
# Exterior 正锥范畴与 Pfaffian 正性搜索设计

- 日期:2026-07-29
- 状态:已按讨论形成书面规格,等待实现前复核
- 负责人:籼至(GitHub `xianzhipan`,Codex 协助)
- 计划实现分支:`work/xianzhi/exterior-positive-category-search`

## 1. 为什么启动这一轮

现有项目已经证明 TN、odd monomial/block-TN 和 tensor-square 三套 determinant
恒正构造,但成功的物理映射都已知或可约化。ZiboJin 的 symmetric-oddcycle 连续族
虽然已经有任意深度严格证明和五模相互作用 transfer,最新完整共同度量审计又证明整个
区间属于已知的 Wei real indefinite-metric contraction 半群。这个结果提供了两条直接
经验:

1. 只枚举另一个固定群、固定度量或孤立矩阵,极易再次命中已知类;
2. exterior-grade 的正锥、尾部支配和连续区间证书确实能组成任意深度定理模板。

因此这一轮不再问“还有哪个群名没扫”,而搜索两个更大的对象:

- **A 主线:exterior-grade domination 与有类型正锥范畴**;
- **B 支线:Pfaffian-TN / positive-spinor 正性**。

目标既可以是一类新的原始正性矩阵,也可以是一条把 TN、外幂锥、Pfaffian 或已知半群
串联起来的更大“矿脉”。有限深度零反例只算候选,不算发现。

## 2. 数学对象

### 2.1 Exterior 分解

对数守恒费米子历史

```text
P = B_L ... B_1
```

使用恒等式

```text
det(I+P) = sum_(k=0)^n tr[(Lambda^k B_L)...(Lambda^k B_1)].
```

`Lambda^k B` 的矩阵元是 `B` 的 `k` 阶 minors。于是每个粒子数 grade 都可看成
subset-state 图上的闭路和。

搜索不要求每个 grade 都逐项正。允许:

- 一部分 grade 在显式有理锥中保持正;
- 一部分危险 grade 有可乘的范数或分块上界;
- 一个严格正的主导 grade 给出下界;
- 短词由精确穷举闭合,长词由 block/tail 不等式闭合。

候选只有在得到

```text
positive lower bound - sum(dangerous-grade upper bounds) > 0
```

的任意深度证书后,才能升级为定理。

### 2.2 有类型正锥范畴

固定一个共同正基只是单对象情形。本轮允许每个 Trotter 位置使用一个标号为 `a` 的
exterior/Fock 正锥 `C_(k,a)`,而时间片是

```text
Lambda^k B_(a->b): C_(k,a) -> C_(k,b)
```

的正映射。实际辅助场语言必须形成合法的闭合路径

```text
a_0 -> a_1 -> ... -> a_L = a_0.
```

只对这种已声明的周期 schedule 证明正性;不把“受限语言恒正”误写成“任意词恒正”。
单一 TN/gauge-TN 锥和固定 semigroup 都是这个结构的退化特例。

闭路“保持一个锥”本身不足以推出普通矩阵 trace 非负。每个被使用的 grade 必须额外满足
以下二者之一:

1. 对每个对象给出显式 simplicial chart `S_(k,a)`,使每条 typed edge 的坐标矩阵
`S_(k,b)^(-1) Lambda^k(B_(a->b)) S_(k,a)` entrywise nonnegative;闭路回到同一
chart 后,对角元和 trace 非负;
2. 给出一个在复合下封闭、对所有允许闭路取非负值且与普通 exterior trace 精确相等的
trace-positive functional。

第一批只实现方案 1;方案 2 只有在先写出精确 functional identity 后才开放。

### 2.3 Pfaffian / Spin 支线

对带 pairing 的 Gaussian operator,不只检查 determinant。把 operator 提升到
full-Fock/Spin 表示后,其矩阵元由 sub-Pfaffians 控制。本轮搜索:

```text
一个固定或有类型的 Fock sign gauge
+ 每个允许时间片在相应 parity block 中逐项非负
+ 周期闭路的 Spin/Fock trace 非负。
```

首轮只做 `2–4` 个复费米模式,以 exact arithmetic、符号可满足性和小规模 SDP 为主。
输入冻结为反对称 Majorana kernel `A^T=-A` 和实际
`U(A)=exp(gamma^T A gamma/4)` Spin lift;固定 Majorana orientation、Fock 基顺序、
even/odd parity trace 与从恒等元连续延拓的 lift 分支。任何只验证
`det(I+D)=Tr_Spin(U)^2` 而没有固定 trace 符号的实现都不验收。

每个物理时间片还必须显式携带整体标量/HS 测度 `c(A)`。实际闭路权重按

```text
w(history) = [product_l c(A_l)] Tr_Spin[U(A_L)...U(A_1)]
```

验收;Spin 双覆盖的连续分支不能替代 `c(A)` 的符号或复相位。只有整个允许历史的
标量乘 Spin trace 为实且非负,才算物理正性。

## 3. 候选语法

### A1. 稀疏 typed exterior-domination edges

- 单粒子维数先取 `n=4,5,6`,只有幸存家族再升到 `7,8`;
- 每个 typed graph 先取 `2–3` 种边,四种边只用于幸存结构;
- 非零有理系数优先来自
`{-2,-1,-1/2,1/2,1,2}`,另允许有理对角尺度;
- support 取有向路径、稀疏 chord、分块反馈和转置配对,但不复刻
symmetric-oddcycle `B(p,q,r)` 语法;
- 每个候选必须有至少两个 cone/chart objects,正性依赖合法的 `a->b` 边和闭合
schedule;若全部对象能合并成一个共同锥或共同 metric,立即降为校准结果;
- types 必须本质不可删除:忘掉 source/target labels 后,untyped alphabet 必须有一个
满足 `det(I+P)<0` 的精确 forbidden word,或有解析证明说明自由词语言的**完整
determinant** 并非恒非负;某个 exterior grade 为负、共同 cone/metric 失败或语言
不闭合都不够。否则它只是给普通 joint alphabet 加标签,不进入本轮候选;
- 优先寻找“某些外幂阶有 typed 正锥、其余阶被主导”的结构,而不是要求原矩阵 TN。

### A2. Support-aware positroid 与多 chart

- 只约束 Cauchy–Binet 展开中实际可达的 minors;
- 允许未被任何合法闭路使用的 minors 为负;
- 用 signed permutation、有理 simplicial cone 和小型网络 chart 作为第一批基变换;
- 通过 `GF(2)`/SAT 检查边符号能否在各对象之间一致消去;
- 若所有 chart 可合并成一个固定 gauge-TN 基,则立即归入已知类。

### A3. 已知类校准

TN、tensor-square、oddcycle fixed/common-metric 家族、所有 untyped joint
`B(p,q,r)` alphabets、single/block exterior CQLF 或 coupled-tail automata,以及简单
sign-regular 两阶段乘积,只用作控制或统一结构校准,不作为新候选计数。ZiboJin 当前
领先的 untyped pair `{p=0.3,p=2.5}` 已完整检查全部 depth-12 words 和十万条
depth-40 随机词;它仍是缺任意深度证书的有限深度候选,归其分支继续推进。

### B1. Pfaffian-TN / positive-spinor

- `2–4` 个复模式;
- 主搜索必须包含真实 pairing;数守恒样本只作实现校准,不能用 determinant 平方代替
Spin trace;
- 枚举稀疏实/复反对称生成元、固定 Fock gauge 和两 chart 交替;
- 对 sub-Pfaffian support 做与 A2 相同的可达闭路筛选;
- 简单 planar matchgate/Ising、标准 positive orthogonal Grassmannian 和
type-D total-positive 子类只作校准。

## 4. 已知类前置排除门

任何候选在昂贵长词搜索前依次检查:

1. 偶 flavor 平方、复共轭模平方或显式 block square;
2. ordinary TN、固定 sign-gauge TN、positive monomial/block-TN;
3. 共同可交换代数、共同 invariant block 和静态扇区直和;
4. 固定 Kramers、split-orthogonal invariant form;
5. **完整共同度量可行性**:先在一粒子空间、再在 `2n` Majorana 表示上,对整个
alphabet 同时求解 Wei/Majorana 型 contraction inequalities,不能只测试一个
预选 metric;
6. tensor/exterior lift 的标准谱恒等式;
7. typed charts 是否只是时间依赖换基/coboundary,并且在具体物理 slice family 上
望远镜约化到某个已知固定锥;只在得到显式约化时归入该已知类;
8. 明确记录它作为抽象证书与 constrained switching、automata-constrained products
和 path-complete multiple-Lyapunov 框架的关系。有限 typed graph + simplicial
charts 总能提升成对象直和上的固定非负块系统,因此不声称这个抽象框架本身新颖;
可能的新颖性只能来自新的 fermionic/HS 嵌入、物理受限语言或更强的不可约定理;
9. Pfaffian 支线中的 Majorana reflection positivity、Kramers、split-O、
Wei contraction、matchgate/Ising 和标准 type-D total positivity。

数值 SDP 找到严格 margin 可作为已知类正证据;数值 SDP 没找到 metric 只能标记为
“未排除”,不能当成不属于已知类的证明。被提升的候选必须补 exact rational
certificate 或可核查的不可行性证书。

## 5. 第一批扫描与算力

### 5.1 参数轴

```text
dimensions = 4, 5, 6
alphabet sizes = 2, 3
word depths = 2, 4, 8, 16, 32
exact exhaustive = two-edge words through depth 8 when the typed graph permits
stress words = alternating, palindromic, transpose-paired,
beam-search and seeded random
pilot budget = 50,000–100,000 compound/determinant checks
full budget = 2–5 million checks only after pilot acceptance
```

pilot 先验证 product order、吞吐量、内存、exact-replay 比例和候选漏斗;只有预计完整
批次仍低于十分钟/16 GB 才进入 full budget。每个负例立即用 exact rational 或高精度
路径重放;数值近零点进入隔离队列,不参与正性
统计。每个候选族最多保留 `10–20` 个有限深度幸存者进入小有理区间或
interval-arithmetic 鲁棒性测试。深度 `16,32` 只做 beam、谱目标和随机变异压力搜索,
绝不写成穷举。

### 5.2 计算资源

第一轮必须控制在单机约十分钟、16 GB 内;先用结构门和低维 exact 算法降低候选数。
只有出现下列任一条件才提交集群参数扫描:

- 通过前置门的独立候选超过 `10^5`;
- 预计检查量超过 `10^8`;
- interval/exact replay 预计单机超过十分钟;
- 维数提升需要超过 16 GB 内存。

超算只负责扩大已经有结构信号的搜索,不用于替代候选定义和新颖性排重。

## 6. 程序架构与数据流

计划新增以下职责清晰的模块:

```text
oracle/exterior_category_search.py
候选语法、typed schedule 与逐词检查

oracle/compound_cones.py
exterior powers、minors、cone/chart 映射和精确交叉检查

oracle/grade_domination.py
block certificate、危险 grade 范数和主导下界

oracle/spinor_positive_search.py
full-Fock/Spin lift、sub-Pfaffian 与 parity-block gauge

oracle/known_class_filters.py
TN、split/Kramers、共同度量、tensor lift 与可约化性筛选

protocols/exterior-positive-category-v1/
axes、settings、provenance、分片与恢复说明

tests/
单元、精确恒等式、已知正负控制和恢复性回归
```

数据流为:

```text
结构化候选
-> 廉价代数/已知类门
-> 短词 exact/高精度反例搜索
-> typed-closure 与 exterior/Pfaffian 检查
-> 长词压力和区间鲁棒性
-> 任意深度证书尝试
-> Hamiltonian/HS 可达性与新颖性审计
```

大体积 cell 输出写入 gitignored 的
`tracks/qmc/results/no-negative-vibes/exterior-positive-category-v1/`。Git 只提交协议、
聚合摘要、最小反例和确认证书。

## 7. 可恢复运行与错误处理

- 每个 cell 原子写 manifest,记录 commit、seed、候选定义、深度和软件版本;
- 重启时只重算 missing/failed cells,不覆盖已完成记录;
- 矩阵乘积使用缩放/log-determinant 路径,避免 overflow;
- `abs(weight)` 或 SDP margin 接近容差时标为 `ambiguous`,转 exact/高精度重放;
- product order、transpose orientation 和 typed source/target 不匹配立即硬失败;
- SDP infeasible、超时和 solver error 分开记录;
- collector 必须检查轴完整性、重复 cell 和设置一致性后才汇总。

## 8. 测试与验收

实现前先冻结下列回归:

1. `Lambda^k` 实现与直接 minors 完全一致;
2. exterior character 总和与 `det(I+P)` 一致;
3. 乘积顺序和转置 convention 有非交换测试;
4. TN、tensor-square 和 oddcycle common-metric 为正控制;
5. 项目已存精确负词为负控制;
6. 非闭合 typed schedule 被拒绝,闭合 schedule 的 trace 可直接重构;
7. Pfaffian/Spin trace 与小 Fock 空间显式矩阵一致;
8. 每片 `c(A)`、HS measure、Spin-lift 分支和闭路总标量均进入显式小 Fock 权重交叉检查;
9. 中断再恢复与一次性运行得到相同汇总;
10. 每个浮点负例都能被 exact 或高精度重放,否则不进入结论。

## 9. 结果等级与停止条件

| 等级 | 含义 | 可说什么 |
|---|---|---|
| S0 | 有精确负例或结构门失败 | 该候选关闭 |
| S1 | 有限深度幸存 | 只是待证候选 |
| S2 | 任意深度 exact/analytic certificate | 已证矩阵或受限语言正性 |
| S3 | S2 + coboundary/已知物理类严格排除 + 完整 HS/受限语言嵌入 | 新 fermionic 受限语言机制强候选;不声称 automaton 框架新颖 |
| S4 | S3 + 有意义模型、可扩展性和独立文献审计 | 可讨论新无符号物理类 |

若第一批所有幸存者都被共同度量、固定 gauge-TN 或标准 Pfaffian 类吸收,停止当前语法,
不追加同分布样本。若 exterior 与 Pfaffian 两线都没有 S1,则保留精确反例地图,并重新
设计候选语法。只有 S2 以上结果才值得大规模扩维。

“发现一条矿脉”还必须同时满足:存在非零宽度的参数区域、至少两个不对易边、interval
certificate、可推广到更多 typed objects 或模式,以及已知类的精确非包含证书。一个
孤立 S2 矩阵不按矿脉报告。

## 10. 与协作者工作的边界

- ZiboJin 的 `work/zibo/representation-cones` 保留 oddcycle、untyped joint
alphabets、block exterior contraction 与 coupled-tail 证书的原始贡献归属;
- ZiboJin 的 `work/zibojin/tensor-square-phase-diagram` 继续负责 tensor-square
`m=3,4,6,8` 的 ED/DQMC 与相图工作;
- 本轮不复制两条分支代码,不重扫任何 untyped joint `B(p,q,r)` alphabet、single/block
exterior CQLF 或 coupled-tail automaton,也不并行做 tensor-square 相图;
- 本轮只把这些结果作为正控制、已知类门和设计经验,新增代码与结论由籼至分支提交。

## 11. 第一轮交付

1. 可恢复的 `exterior-positive-category-v1` 协议;
2. 2–5 百万次低维筛选的聚合表;
3. 每个失败家族的最小精确反例;
4. 所有 S1 幸存者的 candidate card 与完整已知类筛选表;
5. 至少一次 arbitrary-depth certificate 尝试,成功或失败都记录;
6. 面向非专家的中文报告:我们试了什么、为什么淘汰、剩下什么以及是否值得上超算。

这份规格只冻结研究对象、排重门和第一批预算;不会预先把有限深度幸存者写成新发现。
Loading