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
156 changes: 156 additions & 0 deletions research/gaussian-rounder-exact-period/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,156 @@
# The sharp Gaussian radial budget at exact period

Let `s=x+y` and `d=x-y` for Gaussian integers `x,y`. The ideal
positive half-transform sends the pair to `((s+i*d)/2,(s-i*d)/2)`.
A rounder must map Gaussian pairs to Gaussian pairs, preserve `s`
and have Euclidean pair error at most one. Its forward outward slack
is the supremum of `max(0, ||F(z)||-||z||)` over Gaussian pairs `z`.

**Theorem.** A rounder of finite global least period `m` exists exactly
when `4` divides `m`. For each such `m`, let `R` be the least odd
integer at least five that does not divide `m`. The minimum forward
outward slack at exact least period `m` is

```text
beta_R = sqrt((R^2+2R+3)/2) - sqrt((R^2+1)/2).
```

The same construction has absolute radial change and inverse outward
slack at most `beta_R`.

## Starting map and lower bound

Ali Romman's PR #7714 supplies two premises for rounders satisfying
the prescribed iterate identity `F^m=id`. It rules out such a map
unless `4` divides `m`, and in the divisible case proves the lower
bound `beta_R` and constructs a total rounder `F_0` attaining it
with absolute radial change at most `beta_R`. Since a map of least
period `m` satisfies the iterate identity, the necessity and lower
bound carry over. We use the stated results of #7714 directly.
[Finite iterate proof](https://github.com/aliromman/integer-mult-bounds/blob/ba5e006ed70119b23d6ef9e2e8627e43a3f4f18b/research/gaussian-finite-iterate-optimum/PROOF.md)

Only the sum-one fiber will change. Use the doubled half-grid
coordinates `(X,Y)` of #7714, with source color

```text
A = {(X,Y) : X and Y odd, X+Y = 0 mod 4}.
Q(X,Y) = (Y,-X).
```

For difference `d=d_r+i*d_i`, the coordinate vector is
`(d_r-d_i,-d_r-d_i)`. This is a bijective coordinate change on
the sum-one fiber. The pair squared norm at `(X,Y)` equals
`1/2+(X^2+Y^2)/4`. An edge in these coordinates has error one
exactly when undoing `Q` on its output makes that output an
axial neighbor of the input at distance two. For example, the
inverse coordinates give `d_r=(X-Y)/2` and `d_i=-(X+Y)/2`.
The ideal edge sends the doubled coordinates to `Q(X,Y)`.
Changing the output by a vector of length two therefore changes
the difference by length sqrt(2) and the pair by length one.

On the tail rings with doubled radii `(R+4j,R+4j+2)` for
`j >= 0`, the predecessor uses `g_0=Q M`. Here `M` is the
annular matching adapted from Jerry Chen's PR #7570. It commutes
with `Q` and satisfies `M^2=id`, so `g_0^2=Q^2=-id` on these
rings. Every orbit there has length four.
[Annular construction](https://github.com/CubeJerry/integer-mult-bounds/blob/053c42c68165cf24fd4e8bc33c2b20d6ac3d399b/research/sharp-order-four-gaussian-rounding/proof/PROOF.md)

## A column of tangential pairs

Write `q=m/4` and set

```text
a = R+4(R(2q+4)+1)
y_k = a+2k, 0 <= k < q
x_0 = a-4, x_1 = a-2.
```

Each ring of radius `y_k` has the top-side pair
`(x_0,y_k) <-> (x_1,y_k)` in the predecessor matching. To see
this directly, on an inner ring of doubled radius `y` the
ordered top-side list runs from `-y+2` to `y-2` by twos. On an
outer ring it runs from `y` to `-y+4` by minus twos. Pair
consecutive entries starting at the first entry. For `y=y_k`,
the horizontal pairs in increasing order start at residues
`a mod 4` in either case. Both selected coordinates are in
the corresponding list, since `y >= a`, and `x_0 = a mod 4`.

Let `p_k` be the point of this pair in `A`, and `u_k` its mate.
The orbit is `(p_k, Q u_k, -p_k, -Q u_k)`. These are `q`
distinct orbits because their points stay on rings of distinct
radii. The points `p_k` and `-p_k` are on opposite sides of
the ring and both belong to `A`.

## Splice the cycles

For each `k=0,...,q-2`, exchange the successors of
`p_k,p_(k+1)` when `k` is even. When `k` is odd, exchange the
successors of `-p_k,-p_(k+1)`. Consecutive switches use opposite
sides of their shared ring. Thus every changed source has its
original successor up to the time that successor is exchanged.

Two odd horizontal coordinates separated by two have opposite
residues modulo four. The source therefore changes horizontal
coordinate from one top-side pair to the next. Before applying
`Q`, the two exchanged successors are consequently vertical
neighbors of the new sources. Negating the pairs preserves this
fact on the bottom side. Every new edge has Gaussian error one.
The edited permutation remains in the sum-one fiber.

Exchanging the successors of two points in distinct cycles
merges those cycles. Starting at the first changed point, the
new orbit follows the second cycle until its changed point,
then follows the first cycle back to the start. In increasing
order of `k`, switch `k` joins the orbit containing rings
`0,...,k` to the untouched cycle on ring `k+1`. This produces
one cycle of length `4q=m` and leaves every other orbit of
`F_0` intact. For `q=1` the selected orbit already has length
four. The full new map has `F^m=id` because every untouched
orbit already has length dividing `m`. The constructed orbit
forces the least period to be exactly `m`.

For example, at `m=8` the first incompatible odd radius is
`R=5`. The splice on rings 169 and 171 gives the cycle

```text
(167,169), (171,-167), (-165,-171), (-171,167),
(165,171), (169,-165), (-167,-169), (-169,165).
```

## Radial cost of a switch

Rotation and negation preserve norm, so each new edge has
the absolute radial change of a vertical neighbor pair with
horizontal magnitude `x >= a-4` and vertical magnitudes
`y,y+2` for `a <= y < a+2q`. Rationalizing gives

```text
(y+1)/(sqrt(1/2+(x^2+(y+2)^2)/4)+sqrt(1/2+(x^2+y^2)/4))
<= (y+1)/sqrt(x^2+y^2)
<= (a+2q)/(sqrt(2)(a-4))
< (1+1/(4R))/sqrt(2).
```

The second inequality uses the actual inner ordinate
`y <= a+2(q-2)` of a switch. For the last inequality,
`a-4 > 4R(2q+4)`.

The bound on these edges is strictly below `beta_R`. In fact

```text
sqrt(2)*beta_R - 1
= 2/(sqrt((R+1)^2+2)+R+1) - 1/(sqrt(R^2+1)+R)
> 2/(2R+3) - 1/(2R)
>= 1/(4R).
```

Here `sqrt((R+1)^2+2) < R+2` and `R >= 5`. The old edges
have absolute radial change at most `beta_R` by #7714. The
lower bound from the iterate theorem forces equality for the
new map's forward outward slack. The predecessor's attaining
cycle at radii `R,R+2` is outside the splice and supplies
equality for the absolute and inverse outward bounds as well.

This proof uses the exact-sum error-one contract and one mixed
sum fiber. A physical caller, tape and stock price, and any use
in a complete multiplication bound remain open.
65 changes: 65 additions & 0 deletions research/gaussian-rounder-exact-period/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
# Exact least period for Gaussian-pair rounding

**Scoped proof.** An exact-sum Gaussian-pair rounder with Euclidean pair
error at most one can have finite global least period `m` exactly when
`4` divides `m`. Its sharp forward outward slack at that least period is
`sqrt((R^2+2R+3)/2) - sqrt((R^2+1)/2)`, where `R` is the least odd
integer at least five that does not divide `m`. The same construction
attains that bound for absolute radial change and inverse outward slack.

The finite iterate theorem supplies the lower bound and a map whose
period divides `m`. The new construction joins `m/4` four-cycles on
consecutive rings in the sum-one fiber into one cycle of length `m`.
The switches alternate between opposite sides of the rings so each
changed edge still has error one. Moving the switches far enough
toward a corner keeps the new edges below the radial bound.
For instance, period eight attains `sqrt(19)-sqrt(13)`.
[Proof](PROOF.md)

## Reproduce

Run from the repository root. The local checks used CPython 3.12.15
and only the standard library.

```sh
pkg=research/gaussian-rounder-exact-period
python3.12 -B scripts/research_package.py --package "$pkg"
python3.12 -B scripts/check_research_commit.py --package "$pkg" --check-source-history
python3.12 -I -W error -B "$pkg/verify.py"
python3.12 -I -W error -O -B "$pkg/verify.py"
```

Both verifier modes print `PASS_GAUSSIAN_ROUNDER_EXACT_PERIOD`.
They check 20 orbits through period 13,860, including 16,940 Gaussian
edges and 8,430 switched successors. The four controls omit part or
all of the splice, reuse a source, or move the splice near the axis.
The written proof covers arbitrary admissible periods. The verifier
checks only the new construction on its stated fixtures.

The contribution is a local precision result. Physical
implementation, caller pricing and any change to the conditional
multiplication bound remain open. The predecessor theorem is a cited
premise. Its suite, the selected cold replay, Lean and Docker were
unrun. Independent human and formal proof-kernel review are unclaimed.

## Open next questions

Which physical caller, if any, needs a prescribed least period?
Identify one caller and its row and address schedule. That would
give a concrete setting for pricing the local splice predicate,
coordinate arithmetic and workspace restoration.

## Credit

This contribution was produced by Nick Petalas. Ali Romman's
finite iterate optimum supplies the lower bound and the starting
permutation. Jerry Chen's annular matching supplies the geometric
construction it uses. Ali Romman's ring classification supplies
the lower-bound framework, and his exact-sum contract supplies the
Gaussian transform. [#7714 at ba5e006](https://github.com/aliromman/integer-mult-bounds/tree/ba5e006ed70119b23d6ef9e2e8627e43a3f4f18b/research/gaussian-finite-iterate-optimum),
[#7570 at 053c42c](https://github.com/CubeJerry/integer-mult-bounds/tree/053c42c68165cf24fd4e8bc33c2b20d6ac3d399b/research/sharp-order-four-gaussian-rounding),
[#7699](https://github.com/CrocSwap/integer-mult-bounds/pull/7699),
[#6509](https://github.com/CrocSwap/integer-mult-bounds/pull/6509).

Apache-2.0. [Task](TASK.json) · [Source provenance](SOURCES.json) ·
[Advance claim](https://github.com/CrocSwap/integer-mult-bounds/discussions/1166#discussioncomment-19718444)
18 changes: 18 additions & 0 deletions research/gaussian-rounder-exact-period/SOURCES.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
{
"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"
}
]
}
95 changes: 95 additions & 0 deletions research/gaussian-rounder-exact-period/TASK.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,95 @@
{
"schema_version": 1,
"task_id": "formal-interface-lemma",
"priority_revision": 8,
"base_commit": "e54523517d0f03e58352e647c403a5af3e525a57",
"original_card": {
"id": "formal-interface-lemma",
"question": "Can one uncovered interface implication be formalized with exact caller and certificate bindings?",
"starting_points": [
"docs/research/research-family-status.md",
"certificates/selected-result.json",
"research/gate-operand-parity-lean/README.md",
"research/staged544-certified-endpoint/README.md",
"research/unconditional-batched-crt/README.md"
],
"methods": [
"Formalize a missing boundary lemma identified by the selected review.",
"Audit a formal theorem-to-certificate binding independently."
],
"acceptance_gate": "State the exact formalized proposition and remaining assumptions; avoid duplicate arithmetic formalizations already present.",
"first_step": {
"command": [
"python3",
"-B",
"scripts/research_selected_baseline.py"
],
"expected": "PASS_RESEARCH_BASELINE_BINDINGS_ONLY",
"inspect": "Resolve the current selected review, separate complex/ordinary BIT/admitted BIT/outer budgets, and compare the card against live handoffs before starting. Historical ceilings require exact source/model equivalence.",
"scope": "Source-binding orientation only; not a scientific replay or all-size proof. Materialize selected inputs if this checkout is sparse."
},
"first_experiment": "Write one missing statement with quantified hypotheses, actual caller and existing coverage. A compact-switch lemma is a candidate only after its mathematical contract is reviewed. Establish a bounded proof plan before a separately budgeted proof-assistant build; do not rerun large Lean jobs as orientation.",
"bottleneck_link": "A missing interface proof can reduce the conditional gap even if it does not change kappa.",
"allowed_variations": [
"Narrow to one formal lemma or one theorem-to-certificate binding."
],
"scope_change_triggers": [
"Formalizing already-covered arithmetic and claiming the full multiplication theorem."
],
"overlap_terms": [
"hyperplane end",
"formal interface",
"Lean binding",
"3807",
"3857",
"4100",
"gate-operand-parity-lean",
"staged544-certified-endpoint",
"unconditional-batched-crt"
],
"existing_checks": [
{
"scope": "PR #3807: scoped gate-operand parity formalization.",
"evidence": "research/gate-operand-parity-lean/README.md",
"remaining": "Exact theorem-to-caller binding and genuinely uncovered interface lemmas.",
"status": "author-checked",
"source_commit": "4c3530f42a6abcc48e7a6bcffd6c901911dc8269",
"selected_record_sha256": "bb2eaa58647b353f88d0aec2eafb9d30df1be53589231b23e12ceef3e4bad129"
},
{
"scope": "PR #3857: certified real-power moment/endpoint arithmetic.",
"evidence": "research/staged544-certified-endpoint/README.md",
"remaining": "Uniform algorithmic implications, not another endpoint digit.",
"status": "author-checked",
"source_commit": "ef4315b8caa605e47a2515004845eb2d2ed71a86",
"selected_record_sha256": "bb2eaa58647b353f88d0aec2eafb9d30df1be53589231b23e12ceef3e4bad129"
},
{
"scope": "PR #4100: compact-switch formalization proposed as future validation.",
"evidence": "research/unconditional-batched-crt/README.md",
"remaining": "Audit the statement and assumptions before a formal proof campaign.",
"status": "author-checked",
"source_commit": "337d04939571a637aeab54a0f58f54fde811663a",
"selected_record_sha256": "bb2eaa58647b353f88d0aec2eafb9d30df1be53589231b23e12ceef3e4bad129"
}
]
},
"scope_status": "changed",
"actual_scope": "Exact global least period for total exact-sum Gaussian-pair rounding at error one and the sharp radial budget.",
"scope_change_reason": "PR #7714 separates the least-period requirement from the iterate identity. Resolve that local obligation with a written proof before considering any physical caller or formal integration.",
"bottleneck_link": "Local precision result. The finite iterate budget also applies when a local contract requires exact least period. A selected caller and a full price remain open.",
"budget_minutes": 60,
"stop_rule": "Stop after a verified least-period splice at the sharp budget, or report the specific remaining obstruction.",
"outcome": "completed",
"evidence": "PROOF.md gives the all-input construction using PR #7714 at ba5e006ed70119b23d6ef9e2e8627e43a3f4f18b. The exact verifier checks 20 affected orbits, 16,940 Gaussian edges, 8,430 changed successors and four controls.",
"author": "Nick Petalas",
"ai_assistance": "Independent human and formal proof-kernel review are unclaimed.",
"open_questions": [
{
"question": "How could a finite Gaussian rounder with a prescribed exact period be priced in a physical caller?",
"why_it_matters": "The radial budget is a bound on output values, while a caller must also pay for branching, coordinate arithmetic, and restoration.",
"first_step": "Find one caller that uses a prescribed local period and measure its row and address schedule before pricing the local splice predicate.",
"blocked_by": "No selected caller with an exact least-period requirement has been identified."
}
]
}
34 changes: 34 additions & 0 deletions research/gaussian-rounder-exact-period/VERIFY.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,34 @@
{
"schema_version": 1,
"status": "ready",
"python": "3.12",
"requirements": [],
"inputs": [],
"commands": [
{
"argv": [
"{python}",
"-I",
"-W",
"error",
"-B",
"{package}/verify.py"
],
"timeout_seconds": 30,
"checks": "Exact splice edge geometry, Gaussian error, radial costs, orbit length and four negative controls."
},
{
"argv": [
"{python}",
"-I",
"-W",
"error",
"-O",
"-B",
"{package}/verify.py"
],
"timeout_seconds": 30,
"checks": "Exact splice edge geometry, Gaussian error, radial costs, orbit length and four negative controls."
}
]
}
Loading