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
63 changes: 63 additions & 0 deletions research/odd-kernel-coeff-bit-tariff/CERTIFICATE.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,63 @@
{
"claim": {
"odd_baseline_strictly_exceeds_dyadic_on_grid": true,
"odd_fused_strictly_exceeds_dyadic_on_grid": true,
"tables_never_pay_on_this_tariff": true,
"typed_child_xor_counts_identical": true
},
"counts": {
"grid_rows": 198,
"odd_baseline_exceeds_dyadic": 165,
"odd_fused_exceeds_dyadic": 165
},
"j_max": 32,
"kappa_decimal": "0.001672660197076918",
"kappa_fraction": "836330098538459/500000000000000000",
"kernels": [
1,
3,
5,
7,
9,
15
],
"no_new_kappa": true,
"odd_kernels": [
3,
5,
7,
9,
15
],
"schema_version": 1,
"spot": {
"T_fused_j2": {
"1": 102,
"15": 168,
"3": 126,
"5": 145,
"7": 147,
"9": 166
},
"T_word": {
"1": 45,
"15": 64,
"2": 38,
"3": 49,
"4": 32,
"5": 55,
"7": 56
},
"non_monotone": {
"T(1)": 45,
"T(2)": 38,
"T(4)": 32
}
},
"status": "PASS_ODD_KERNEL_COEFF_BIT_TARIFF",
"tariff": {
"definition": "T_word(k) = sum over affine diagonal/control coefficients c=p/q in lowest terms of bit_length(|p|)+bit_length(q); T_fused(q,j)=2*T_word(q*2^j)+T_word(q*2^{j+1}); T_baseline(q,j)=4*T_word(q*2^j).",
"name": "ADDRESS_affine_coeff_num_den_bit_sum",
"unit_coeff_bits": 20
}
}
48 changes: 48 additions & 0 deletions research/odd-kernel-coeff-bit-tariff/MASTER-HANDOFF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,48 @@
# MASTER-HANDOFF — odd-kernel-coeff-bit-tariff

## Claim (one sentence)

Under the named ADDRESS affine coefficient-bit tariff, every certified odd
kernel `q∈{3,5,7,9,15}` strictly exceeds the dyadic fused and baseline bills
on `j=0..32`, so larger-`q` coefficient tables never pay for themselves
despite identical typed child/XOR counts (answers #2687 OQ2).

## Closest prior

- #2687 / `research/odd-kernel-query-pair` — exclusion pattern; OQ2 left open
- PR840 / `research/dyadic-query-pair` — dyadic fusion and ADDRESS shape
- triadic-query-pair — kernel-3 specialization feeding #2687

## New checkable fact

Named tariff `T_word` / `T_fused` / `T_baseline`; closed form; prior
`address_word` cross-check; `165+165` strict odd-over-dyadic exceedances;
non-monotonicity of `T_word` (`32<38<45` at `|k|=4,2,1`).

## Evidence commands

```sh
python3.12 -I -B research/odd-kernel-coeff-bit-tariff/verify.py
python3.12 -I -O -Werror -B research/odd-kernel-coeff-bit-tariff/verify.py
python3.12 -I -B research/odd-kernel-coeff-bit-tariff/independent.py
python3.12 -I -B research/odd-kernel-coeff-bit-tariff/test_package.py
python3.12 -B scripts/research_package.py --package research/odd-kernel-coeff-bit-tariff
```

## Soft-overlap vs leave-alones

Clear of flag-0 tail / stock envelope / gate-operand / C5 / side-correction /
present-head / groups B–D native clusters. Touches only the #840 / #2687
odd-kernel cascade at the unanswered coefficient-bit OQ. Soft-overlap with
#2687 is intentional predecessor pin of `program.py`, not a typed-invoice
reprint.

## Attribution

Author Wentao Li `<wentaol4@illinois.edu>`. Neutral AI assistance disclosure.
No `Co-authored-by: Victor Li`. No CrocSwap PR from this worker (master publishes).

## Disposition

Ready for master independent review / publish decision. Kappa unchanged
(`0.001672660197076918`).
17 changes: 17 additions & 0 deletions research/odd-kernel-coeff-bit-tariff/NOTICE.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,17 @@
# Notice — research/odd-kernel-coeff-bit-tariff

Prepared by Wentao Li (GitHub VictorLiwentao, X @Is__Victor) with substantial
AI assistance. New material is Apache-2.0 under the repository LICENSE.

Predecessor credit:

- Wentao Li / #2687 (`research/odd-kernel-query-pair`) — general odd-kernel
exclusion pattern; OQ2 (coefficient-table bit tariff) answered here.
- Wentao Li / PR840 (`research/dyadic-query-pair`) — three-query dyadic fusion
and ADDRESS lowering shape.
- Pending fork package `research/triadic-query-pair/` — kernel-3 specialization
feeding #2687.
- Karl Appel / applecrusher / PR521 — native unipotent ADDRESS query lowering.
- commdata2338 / PR677 — completed DATA-shear boundaries without a private row.

No independent human review is claimed. No selected frontier or kappa change.
92 changes: 92 additions & 0 deletions research/odd-kernel-coeff-bit-tariff/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,92 @@
# Coefficient-bit tariff for odd-kernel ADDRESS lowerings

## Claim

Let `address_word(k)` be the seven-event ADDRESS lowering of
`M(k)=I+(k/4)sr^T` used by PR840 / `#2687`. Define the **coefficient-bit
tariff**

```text
T_word(k) = Σ_{c = p/q in affine coeffs} ( bit_length(|p|) + bit_length(q) )
```

with each `c` in lowest terms. For kernel `q` and dyadic index `j`,

```text
T_fused(q,j) = 2·T_word(q·2^j) + T_word(q·2^{j+1})
T_baseline(q,j) = 4·T_word(q·2^j)
```

matching the three-query fused and four-query separate power lists. Then for
every certified odd kernel `q ∈ {3,5,7,9,15}` and every `j ∈ {0,…,32}`:

```text
T_fused(q,j) > T_fused(1,j)
T_baseline(q,j) > T_baseline(1,j).
```

Typed child / affine / XOR counts remain `3/18/2` (fused) and `4/24/4`
(baseline) independent of `q`. Therefore, on this named tariff, larger-`q`
coefficient tables never pay for themselves against the dyadic path.

Selected kappa is unchanged.

## Closed form for `T_word`

The affine coefficients of `address_word(k)` are ten units `±1` together with

```text
a, 1+a, -a, -a, -1/a, 1/a - 1 where a = k/4.
```

Each unit contributes `bit_length(1)+bit_length(1)=2`, so

```text
T_word(k) = 20 + bits(a) + bits(1+a) + 2·bits(a) + bits(4/k) + bits((4-k)/k)
```

with `bits(p/q)=bit_length(|p|)+bit_length(q)`. The formula is sign-invariant
(`T_word(k)=T_word(-k)`) and matches direct enumeration on `|k|≤64`.

## Non-monotonicity

`T_word` is **not** monotone in `|k|`:

```text
T_word(4) = 32 < T_word(2) = 38 < T_word(1) = 45.
```

A word-level “larger `|k|` ⇒ larger bill” argument is false. The claim is the
fused/baseline program comparison on the certified kernel grid, not word
monotonicity.

## Grid evidence

For each of the `5×33 = 165` odd pairs `(q,j)` the verifier checks fused and
baseline strict exceedance separately (`165` comparisons each). Spot values at `j=2`:

| q | T_fused(q,2) | T_baseline(q,2) |
|---:|---:|---:|
| 1 | 102 | 128 |
| 3 | 126 | 160 |
| 5 | 145 | 184 |
| 7 | 147 | 188 |
| 9 | 166 | 212 |
| 15 | 168 | 216 |

Fusion saves coefficient bits for every kernel, but the odd fused bill remains
strictly above the dyadic fused bill on the whole grid.

## Relation to `#2687` / PR840

`#2687` proved the excluded-prime admission pattern and recorded identical
typed invoices for all odd `q`, leaving open whether coefficient-table
bit-sizes can still favor some `q` under a concrete physical tariff. This
package attaches one named tariff to the same `address_word` and answers that
obligation in the negative on `j=0..32` for the certified kernels.

## What this does not claim

- No alternate tariff (max-height only, packing, modulus-dependent encoding).
- No BIT-supplier hosting bill (still `#2687` OQ1).
- No new kappa and no production adoption.
83 changes: 83 additions & 0 deletions research/odd-kernel-coeff-bit-tariff/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,83 @@
# Odd-kernel coefficient-bit tariff (answers #2687 OQ2)

**Card `connector-alphabet-change`, priority revision 6**, refined to open
question 2 of
[#2687](https://github.com/CrocSwap/integer-mult-bounds/pull/2687)
(`research/odd-kernel-query-pair`). No new kappa
(`0.001672660197076918`).

`#2687` showed that every odd kernel `q` has the same typed fused invoice as
the dyadic case (`4→3` children, `24→18` affine, `4→2` XOR) and left open
whether coefficient-table bit-sizes can still favor some larger `q` under a
concrete physical tariff. This package attaches one named tariff to the PR840
`address_word` lowering and answers that obligation.

**Tariff.** For each affine diagonal/control coefficient `c=p/q` in lowest
terms, charge `bit_length(|p|)+bit_length(q)`. Summing over `address_word(k)`
gives `T_word(k)`. Program bills are

```text
T_fused(q,j) = 2·T_word(q·2^j) + T_word(q·2^{j+1})
T_baseline(q,j) = 4·T_word(q·2^j)
```

**Result.** On the certified kernels `{1,3,5,7,9,15}` and every
`j∈{0,…,32}`, every odd kernel strictly exceeds the dyadic fused and baseline
bills (`165+165` comparisons). `T_word` is not monotone in `|k|`
(`T(4)=32 < T(2)=38 < T(1)=45`), so the claim is the program comparison, not
a word-level size argument. Larger-`q` tables never pay for themselves on this
tariff.

```sh
python3.12 -I -B research/odd-kernel-coeff-bit-tariff/verify.py
```

Expected:

```text
PASS_ODD_KERNEL_COEFF_BIT_TARIFF kernels=[1, 3, 5, 7, 9, 15] j=0..32 odd_fused_exceeds_dyadic=165 odd_baseline_exceeds_dyadic=165 no_new_kappa
```

## Evidence

- Closed-form `T_word` (ten unit coeffs + six variable rationals in `a=k/4`)
matches enumeration on `|k|≤64` and the byte-pinned `#2687` `address_word`.
- Grid: odd fused/baseline bills exceed dyadic at every `(q,j)` above.
- Spot `j=2`: dyadic fused `102` vs odd `{126,145,147,166,168}`.
- Fusion saves bits for every kernel but does not overturn odd `>` dyadic.
- Non-monotonicity control and inflated-dyadic adversarial control.

Default verify is read-only against `CERTIFICATE.json`. Admission uses
`need()` and stays enforced under `python -O`. Explicit `--write` regenerates
the certificate only.

## Reproduce

```sh
python3.12 -I -B research/odd-kernel-coeff-bit-tariff/verify.py
python3.12 -I -O -Werror -B research/odd-kernel-coeff-bit-tariff/verify.py
python3.12 -I -B research/odd-kernel-coeff-bit-tariff/independent.py
python3.12 -I -B research/odd-kernel-coeff-bit-tariff/test_package.py
python3.12 -B scripts/research_package.py --package research/odd-kernel-coeff-bit-tariff
```

## Open next questions

1. **Does any alternate physical tariff ever make a certified odd kernel beat
dyadic at some `j`?** First step: name max-height / packing / modulus-dependent
encoding on the same coefficients and replay the grid. Blocked by: no
alternate tariff admitted here.
2. **Can an admitted BIT supplier host a small odd-kernel shear with a favorable
fully charged bill versus dyadic?** First step: one exact source-return
boundary with routing, precision and dispatch. Blocked by: still `#2687` OQ1;
not constructed here.

## Credit

Credit `#2687` / `research/odd-kernel-query-pair` (OQ2 answered), PR840 /
`research/dyadic-query-pair`, the pending triadic kernel-3 package, and
PR521/PR677 ADDRESS / DATA-shear lineage. Predecessor notices are unchanged.

Prepared by Wentao Li ([VictorLiwentao](https://github.com/VictorLiwentao),
[@Is__Victor](https://x.com/Is__Victor)) with substantial AI assistance, under
Apache-2.0. No independent human review is claimed. No new kappa.
48 changes: 48 additions & 0 deletions research/odd-kernel-coeff-bit-tariff/SOURCES.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,48 @@
{
"schema_version": 1,
"base_commit": "20ecf2891d80b5acf18d980b2164fb44ea7faa90",
"files": [
{
"path": "research/odd-kernel-query-pair/program.py",
"usage": "input",
"git_blob": "93699aec204ee787418254b10e50a1dd84e8bfc3",
"sha256": "f167826e66d39cd448651f7cdb27ba150a846786b2e1a9a2c7dad5100028a7c2",
"size_bytes": 6332
},
{
"path": "research/odd-kernel-query-pair/README.md",
"usage": "provenance",
"git_blob": "5b79bbd87534804fff944ca8de5e7ce88a07e9c9",
"sha256": "b7c1cf0518eff8cc7c9682e9038cb78d93ce9cae7b2d44f40a8286875fb019a2",
"size_bytes": 4258
},
{
"path": "research/odd-kernel-query-pair/TASK.json",
"usage": "provenance",
"git_blob": "1834f1fb4e7727708f9c89104fff50cd48281332",
"sha256": "886925d67cd1f52dc00591574217751bbe9bffd82956c6cefbc35267b32c8dfb",
"size_bytes": 4576
},
{
"path": "research/dyadic-query-pair/PROOF.md",
"usage": "provenance",
"git_blob": "aaba3e501de1f761c783ec7f0dc9307ef45ab5d7",
"sha256": "9a5ce93b866371189c797fbe8837c6a0bd22c5cf71ed84d60b9ae1c22a5ec83a",
"size_bytes": 10567
},
{
"path": "certificates/selected-result.json",
"usage": "provenance",
"git_blob": "b1fea0c94436b236cc2b7d41a7be032811b94c81",
"sha256": "bb2eaa58647b353f88d0aec2eafb9d30df1be53589231b23e12ceef3e4bad129",
"size_bytes": 10725
},
{
"path": "docs/research/agent-task-catalog.json",
"usage": "provenance",
"git_blob": "67f085acd86acd7d4e03922d9cf4b26bca66fed5",
"sha256": "47b42433f5a85809a3aadd29c2c339da2c6ed728e4f7762df6b88bccb7874cf0",
"size_bytes": 58553
}
]
}
Loading