Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 18 additions & 1 deletion docs/ci-verification.md
Original file line number Diff line number Diff line change
@@ -1,8 +1,25 @@
# CI verification

## Current selected-family replay: #6030
## Current selected-family replay: #7000 with #7001

The [selected record](../certificates/selected-result.json) binds conditional
κ = 0.00363005817 to the [application review](research/pr-intake-evidence/pr7000-application/PROOF-REVIEW.md),
[validation receipts](research/pr-intake-evidence/pr7000-application/VALIDATION.json),
and [explicit source closure](../certificates/frontier7000-source.json).
`make frontier7000-verify` follows that exact-source selected replay. The review
reuses eleven authenticated preceding executions and records seven fresh serial
application, scalar and complete native-catalogue executions. The earlier
monolithic timeout is retained separately and is not completed coverage.
Consult the receipts for exact commands, source heads and limits.

The five inherited framework assumptions remain. No complete uniform word,
literal multiplier execution, independent human proof or formal-kernel proof
is asserted. Routine research submissions do not need to rerun this
selected-family replay. The sections below preserve historical checkpoints.

## Historical selected-family replay: #6030

The [historical checkpoint](../certificates/checkpoints/frontier6030.json) binds conditional
κ = 0.003125804 to the [proof review](research/pr-intake-evidence/pr6030-audit/PROOF-REVIEW.md),
[validation receipts](research/pr-intake-evidence/pr6030-audit/VALIDATION.json),
and [explicit source closure](../certificates/frontier6030-source.json).
Expand Down
143 changes: 143 additions & 0 deletions research/bit-five-term-residue-bound/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,143 @@
# Five terms still have residue rank at most two

This contribution was produced by Nick Petalas. Apache-2.0.

## Statement and credited premises

Let m >= 4, V = M_m(F2), and let U_j(z) = L_j z R_j for
L_j,R_j in GL_m(F2), j = 1,...,5. Suppose a_j in V has matrix rank
at most one and

sum_j a_j = 0,
sum_j a_j * (U_j z) = 0 in Sym^2(V), for every z in V. (1)

The multiplication in (1) is ordinary polynomial multiplication of linear
forms. Squares are retained. Define beta = sum_j U_j^-1(a_j).

**Theorem.** rank(beta) <= 2. For every m >= 4 there is a solution with
five nonzero rank-one a_j and rank(beta) = 2.

The model and necessary conditions are JoyboyBrian's #5490, Sections 1-3
of `research/bit-three-term-twist-rigidity/PROOF.md`. We cite its
transport-distance lemma: distinct U and U' have rank(U+U') >= m as
linear maps on V. We also cite Nick Petalas's #5978, Theorem parts 2-3
in `research/bit-four-term-twist-escape/PROOF.md`: for four contributions
at m >= 3, a nonzero residue requires all four frames equal to x and
sum U_j = 0; its rank is at most 2 rank(x), and rank two is attained by
rank-one x at every m >= 3. These are premises, not new verification
claims. SOURCES.json pins the exact proof bytes on canonical main.

The new step is the complete reduction from five contributions to that
four-contribution theorem. It answers the m >= 4 rank-one part of
Open next question 2 in #5978. It does not answer its proposed m = 2 census
or the exceptional five-term m = 3 case.

## Linear-syzygy fact

Use the same linear-syzygy fact as #5978. If independent linear forms
b_1,...,b_d and linear forms f_1,...,f_d satisfy sum b_i f_i = 0,
then f_i lies in span{b_h : h != i}. To see the needed implication,
extend the b_i to a coordinate basis of the polynomial ring. The
coefficient of b_i^2 makes the b_i coefficient of f_i zero. The
coefficient of b_i c for each complementary coordinate c makes that
coefficient of f_i zero. Terms b_i b_h allow only the other b_h.
This explains why discarding squares would change the argument.

Consequently, if sum b_i (F_i z) = 0 for all z, then the linear map
F_i has image dimension at most d-1. All ranks in this paragraph are
vector-space ranks on V, rather than the matrix rank of a frame.

## Reduction by frame-span dimension

If one a_j is zero, delete that term. At most four terms remain and the
credited four-term theorem applies after adding zero terms if necessary.
Its nonzero-residue case has four equal rank-one frames, so rank(beta)
is at most two. Hence assume all five a_j are nonzero.

Put d = dim span{a_1,...,a_5}. Their zero sum gives d <= 4.
Dimension one is impossible: over F2 five equal nonzero vectors sum to
that same nonzero vector. The remaining cases are as follows.

**d = 4.** After reordering, the frames are independent x_1,...,x_4
and x_1+...+x_4. Substitution into (1) gives

sum_(i=1)^4 x_i * ((U_i+U_5) z) = 0.

Each U_i+U_5 has image dimension at most three by the syzygy fact.
Since 3 < m, the credited distance lemma forces U_i = U_5 for every
i. Thus beta = U_5^-1(sum a_j) = 0.

**d = 3.** The frames, after reordering, are x,y,x+y,z,z with x,y,z
independent. Here is the full multiplicity argument. The set of values
occurring an odd number of times has odd size at most five and sums to
zero. Size one is impossible. Size five is impossible in F2^3: all
seven nonzero vectors sum to zero, so its two-element complement would
also sum to zero, despite having distinct elements. Thus there are
three odd-multiplicity values x,y,x+y. The other two entries are an
equal pair z,z. If z lay in span{x,y}, d would be at most two.

Equation (1) becomes

x*((U_1+U_3)z') + y*((U_2+U_3)z') + z*((U_4+U_5)z') = 0

for every argument z' in V. Each displayed difference has image
dimension at most two, less than m. Therefore U_1=U_2=U_3 and
U_4=U_5. The triple and the pair cancel separately in beta, giving zero.

**d = 2.** The three possible nonzero frame values are x,y,x+y. Their
multiplicities all have the same parity because their sum is zero.
Their total is five, so each multiplicity is odd; the only possibility
is 3,1,1. Reorder the frames as x,x,x,y,x+y. Equation (1) reads

x*((U_1+U_2+U_3+U_5)z) + y*((U_4+U_5)z) = 0.

The second map has image dimension at most one. Transport distance
forces U_4=U_5. The equation then forces U_1+U_2+U_3+U_5=0, since
multiplication by the nonzero linear form x is injective. Moreover

beta = (U_1^-1+U_2^-1+U_3^-1+U_5^-1)(x).

These are exactly four equal-frame contributions satisfying (1).
The cited four-term theorem gives rank(beta) <= 2 rank(x) <= 2.
This proves the bound in all cases.

## Attainment with five nonzero rank-one frames

Take a credited four-term rank-two witness at this m, with all four
frames x = u v^T and invertible transports U_1,...,U_4 summing to zero.
Choose w independent of u and replace the first contribution by

(w v^T, U_1), ((u+w) v^T, U_1).

Both new frames have rank one, are nonzero, and sum to x. This
replacement preserves each equation in (1) and the residue exactly.
There are now five nonzero frames and residue rank two. No new
four-term construction is claimed.

For direct finite corroboration, identify F2^m with F_(2^m) and write
M_c for multiplication by c. The checked new five-term witness has

frames = (e_1 e_0^T, (e_0+e_1)e_0^T, x, x, x), x=e_0 e_0^T,
left multipliers = (1, 1, b, 1+b, 1+b),
right multipliers = (1+b, 1+b, 1+b, 1, b),
U_j(z) = M_left[j] z M_right[j], b = t.

The finite verifier checks all input basis matrices in (1), inverse
transport, all five frame ranks and the actual rank of beta at
m=4,5,6,7,8. The moduli are t^4+t+1, t^5+t^2+1, t^6+t+1,
t^7+t+1 and t^8+t^4+t^3+t+1. It checks a multiplicative inverse for
every nonzero element of each finite field. These new five-frame checks
do not replay the predecessor's census or its four-frame verifier.

## Scope and next step

This is a necessary-condition obstruction in one off-diagonal block.
A five-contribution block with rank-one frames cannot alone supply a
full-rank residue when m >= 4. Other blocks, direct diagonal contributions,
the second-order constant-input cross terms and complete words may alter
a caller's total target. No arbitrary-word lower bound follows.

The threshold m >= 4 is used at d=4, where the rank-three image bound
need not identify the transports if m=3. This case is left open. Native
twist costs, physical realization, recursive invoices, the five selected
framework assumptions and a complete multiplication theorem are unrun.
110 changes: 110 additions & 0 deletions research/bit-five-term-residue-bound/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,110 @@
# Five address-twist contributions still have residue rank at most two

This contribution was produced by Nick Petalas. Apache-2.0.

**Claim type: scoped written proof with exact finite corroboration.** For
every m >= 4, five rank-at-most-one frames satisfying the dirty degree-one
conditions of JoyboyBrian's [#5490](https://github.com/CrocSwap/integer-mult-bounds/pull/5490)
have role-s residue of matrix rank at most two. The bound is attained with
five nonzero frames at every such m.

Let V=M_m(F2), U_j(z)=L_j z R_j with invertible L_j,R_j, and require

```text
sum a_j = 0,
sum a_j * (U_j z) = 0 in Sym^2(V), for every z,
beta = sum U_j^-1(a_j).
```

Polynomial squares are retained. [PROOF.md](PROOF.md) classifies the
five-frame span and multiplicities. At span dimension four or three,
transport rank distance forces residue zero. At dimension two, the frame
multiplicities are 3,1,1 and the problem reduces to the credited four-term
theorem of [#5978](https://github.com/CrocSwap/integer-mult-bounds/pull/5978).
Its rank-two witness becomes a five-frame witness by splitting one frame.

This answers the m >= 4 rank-one part of that older package's Open next
question 2. It does not settle five contributions at m=2 or m=3, arbitrary
frame ranks, complete words, supplier admission or a new kappa. In this
necessary-condition model, adding a fifth contribution does not raise the
maximum residue rank above two at m >= 4.

## Evidence and reproduction

From the repository root, standard-library CPython 3.12:

```sh
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -W error -B research/bit-five-term-residue-bound/verify.py
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -W error -O -B research/bit-five-term-residue-bound/verify.py
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -B scripts/research_package.py --package research/bit-five-term-residue-bound
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -B scripts/check_research_commit.py --package research/bit-five-term-residue-bound --check-source-history
```

The scientific marker is `PASS_FIVE_TERM_RESIDUE_BOUND`. The verifier
enumerates all five-nonzero-vector multisets spanning F2^d and summing to
zero for d=1,2,3,4. Their counts are 0,3,28,168. It checks the claimed
types and every kernel-basis syzygy for each of the 199 multisets in a
six-dimensional ambient space. This finite calculation corroborates the
written dimension-independent argument; it does not replace that proof.

It then checks new five-frame rank-two witnesses at m=4,...,8 by direct
polynomial arithmetic on every input basis matrix. Omitted-term,
changed-frame and changed-transport controls must fail at the named
conditions; a square-retention control is also required. No predecessor
executable is imported or replayed. The two cited proof files are
authenticated by their pinned hashes.

Default replay writes nothing. Its JSON output includes the complete
finite result and SHA256 of all six package files and the two proof inputs.
`--receipt /absolute/path/outside/repository.json` writes the same generated
receipt and refuses to overwrite an existing file. No receipt is hand-written.

## Limits and credit

The universal result is a written proof using cited premises. Independent
human and formal proof-kernel review are unclaimed. No rounded native
implementation, complete caller, other off-diagonal blocks, constant-input
second-order cross terms, physical twist costs, recursive moment or selected
supplier replay is supplied. The selected frontier is unchanged.

JoyboyBrian's #5490 supplies the model, the necessary conditions, transport
rank distance and three-term rigidity. Nick Petalas's older #5978 supplies
the four-term theorem, its field witnesses and the explicit five-term
question. Krzysztof Wójt's #5003 supplies the inherited second-order
quotient and source compatibility, with the Cinna Davis/#2110, Ali Romman,
eumemic and huxint lineage retained by citation. Linear syzygies and finite
fields are standard ingredients. The new contribution is the five-term
reduction, its sharpness by splitting, and the new finite corroboration.

The scoped side result changes card `bit-new-register-model` from a native
alphabet search to this local obstruction. The card's working source-binding
baseline passed against selected #7000. This is not a new register census
or a completed native alphabet. The CLAIM describes the local refinement;
TASK.json marks the change from the broader catalogue premise explicitly.

## Open next questions

1. Can five rank-one frames at m=3 have residue rank three? The span-four
branch is the sole new branch left by this proof. Start with a five-circuit
of rank-one matrices and impose the full quadratic equations on two-sided
transports. A transport pair can differ by rank three there, so the present
distance argument does not settle it.
2. What is the first contribution count admitting full-rank residue for a
named m>=4? Start with six contributions and classify frame multiplicities
before any word synthesis. A local residue still needs every remaining
block, constant-input cross term and physical cost in a complete caller.

Structured relevance, first steps and blockers are in [TASK.json](TASK.json).

## Implementation bounds

The verifier has no recursion, network, subprocesses, dynamic code generation
or predecessor imports. Census loops cover at most C(19,5)=11,628 multisets
per dimension; syzygy systems have 30 variables and 21 equations. Witnesses
have m<=8, at most 64 input basis matrices, five terms and 4,096 polynomial
bit positions. Functions are shorter than 60 physical lines. Explicit
exceptions remain active under optimization. Each registered command has a
120-second timeout. Python exact arithmetic allocates during this bounded
offline computation, a Power of Ten rule 3 deviation. Package and proof
files are individually capped at 256 KiB. Warning-strict normal and
optimized execution provide the interpreted-code diagnostics.
30 changes: 30 additions & 0 deletions research/bit-five-term-residue-bound/SOURCES.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
{
"schema_version": 1,
"base_commit": "e54523517d0f03e58352e647c403a5af3e525a57",
"files": [
{
"path": "certificates/selected-result.json",
"usage": "provenance",
"git_blob": "75d77b328ea02ded535e7f23aee5b1e55d0bb1a8",
"sha256": "154c3d30455d778fd3343b6b4851f0fa4665f0f7625b0adb3675e2bafbbf6bf1"
},
{
"path": "docs/research/agent-task-catalog.json",
"usage": "provenance",
"git_blob": "3404fbdd5e7529a0674eef26114b5071e4586cec",
"sha256": "ead6cdc9d0af6d58e8a49744d8ec8c7c15e13d2ba93c276c434bcf9b51b93a81"
},
{
"path": "research/bit-four-term-twist-escape/PROOF.md",
"usage": "input",
"git_blob": "7893702732f4903adb21e5ec8696cede0aefb5f5",
"sha256": "93ed15ab008347e9a410d6fc306ba75897cab4b4b4d0b0ba5d3fa1a3d78d122f"
},
{
"path": "research/bit-three-term-twist-rigidity/PROOF.md",
"usage": "input",
"git_blob": "65120d64fb7e25765734584a7e389e53e83fa142",
"sha256": "6ca59361eefe1fa546d348aefce127b325149323a83d4ab10445ac6f28c6a854"
}
]
}
Loading