diff --git a/research/relay1256-sharp-scalar-range/.gitattributes b/research/relay1256-sharp-scalar-range/.gitattributes new file mode 100644 index 0000000000..3c84e2ff25 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/.gitattributes @@ -0,0 +1,2 @@ +# Keep the source notices byte-exact. +credits/* -text whitespace=cr-at-eol diff --git a/research/relay1256-sharp-scalar-range/CREDITS.md b/research/relay1256-sharp-scalar-range/CREDITS.md new file mode 100644 index 0000000000..251a3eba75 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/CREDITS.md @@ -0,0 +1,43 @@ +# Sources and attribution + +Ali Romman prepared this range proof, exact basis audit, scalar witnesses +and package with substantial OpenAI Codex assistance. Apache-2.0 applies. +The l1 formula for a linear form on a cube is elementary; it is not +claimed as a new general norm theorem. + +Clear1nsight supplied the role-1256 graph and physical analysis in +[PR #7704](https://github.com/CrocSwap/integer-mult-bounds/pull/7704), +head `e29a5ae8591243c1178067727f12ce4abefa0b93`. +The exact [graph](https://github.com/Clear1nsight/integer-mult-bounds/blob/e29a5ae8591243c1178067727f12ce4abefa0b93/research/cf6a-banked-role1256-relay/data/CANDIDATE-GRAPH.json.gz) +is copied without change. Its compressed SHA-256 is +`0177c25371181e3b1f2c13325eb8434fbf6384a9ab99a612849efd5d0f6ab2cc`; +its decompressed SHA-256 is in PROOF.md and in the checker. +The original [physical code](https://github.com/Clear1nsight/integer-mult-bounds/blob/e29a5ae8591243c1178067727f12ce4abefa0b93/research/cf6a-banked-role1256-relay/physical.py) +and [receipt](https://github.com/Clear1nsight/integer-mult-bounds/blob/e29a5ae8591243c1178067727f12ce4abefa0b93/research/cf6a-banked-role1256-relay/evidence/PHYSICAL.json) +give the separate-product triangle bound and the source's physical scope. +They were read, but are not imported by this checker. This package does +not rerun or certify the complete physical application. + +The [source notice](https://github.com/Clear1nsight/integer-mult-bounds/blob/e29a5ae8591243c1178067727f12ce4abefa0b93/research/cf6a-banked-role1256-relay/NOTICE) +identifies SYLforge's nine-swap word in +[PR #7700](https://github.com/CrocSwap/integer-mult-bounds/pull/7700) +as the word source. It also credits DreamingOfClouds, Ali Romman, Rohan +Arun, Jerry Chen (CubeJerry), Douglas Colkitt and the retained caller, +native, bank and selected-framework contributors. The source's notice, +all eight files in its credits directory, and license are preserved here. +Their historical selected checkpoints remain historical. + +Ali Romman's +[PR #7702](https://github.com/CrocSwap/integer-mult-bounds/pull/7702) +gives [row-atomic arithmetic](https://github.com/aliromman/integer-mult-bounds/blob/785860501b03893eaa9f5279f978287e75d3f3af/research/chart595-row-atomic-arithmetic/PROOF.md) +on a different chart-595 graph. It is related work, not a consumed +range bound. It allows independently bounded donors. This package keeps +the role-1256 graph's entrance dependencies and its separate divisions. +Its W therefore has a different meaning. No code from that package is +copied here. + +Douglas Colkitt and the repository contributors supplied the package +workflow and main-branch provenance machinery. SOURCES.json binds the +selected record and task catalogue at the base commit. The unmerged +graph is bound by its explicit source commit and byte hashes above; +it is not represented as an ancestor of main. diff --git a/research/relay1256-sharp-scalar-range/LICENSE b/research/relay1256-sharp-scalar-range/LICENSE new file mode 100644 index 0000000000..261eeb9e9f --- /dev/null +++ b/research/relay1256-sharp-scalar-range/LICENSE @@ -0,0 +1,201 @@ + Apache License + Version 2.0, January 2004 + http://www.apache.org/licenses/ + + TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION + + 1. Definitions. + + "License" shall mean the terms and conditions for use, reproduction, + and distribution as defined by Sections 1 through 9 of this document. + + "Licensor" shall mean the copyright owner or entity authorized by + the copyright owner that is granting the License. + + "Legal Entity" shall mean the union of the acting entity and all + other entities that control, are controlled by, or are under common + control with that entity. For the purposes of this definition, + "control" means (i) the power, direct or indirect, to cause the + direction or management of such entity, whether by contract or + otherwise, or (ii) ownership of fifty percent (50%) or more of the + outstanding shares, or (iii) beneficial ownership of such entity. + + "You" (or "Your") shall mean an individual or Legal Entity + exercising permissions granted by this License. + + "Source" form shall mean the preferred form for making modifications, + including but not limited to software source code, documentation + source, and configuration files. + + "Object" form shall mean any form resulting from mechanical + transformation or translation of a Source form, including but + not limited to compiled object code, generated documentation, + and conversions to other media types. + + "Work" shall mean the work of authorship, whether in Source or + Object form, made available under the License, as indicated by a + copyright notice that is included in or attached to the work + (an example is provided in the Appendix below). + + "Derivative Works" shall mean any work, whether in Source or Object + form, that is based on (or derived from) the Work and for which the + editorial revisions, annotations, elaborations, or other modifications + represent, as a whole, an original work of authorship. For the purposes + of this License, Derivative Works shall not include works that remain + separable from, or merely link (or bind by name) to the interfaces of, + the Work and Derivative Works thereof. + + "Contribution" shall mean any work of authorship, including + the original version of the Work and any modifications or additions + to that Work or Derivative Works thereof, that is intentionally + submitted to Licensor for inclusion in the Work by the copyright owner + or by an individual or Legal Entity authorized to submit on behalf of + the copyright owner. For the purposes of this definition, "submitted" + means any form of electronic, verbal, or written communication sent + to the Licensor or its representatives, including but not limited to + communication on electronic mailing lists, source code control systems, + and issue tracking systems that are managed by, or on behalf of, the + Licensor for the purpose of discussing and improving the Work, but + excluding communication that is conspicuously marked or otherwise + designated in writing by the copyright owner as "Not a Contribution." + + "Contributor" shall mean Licensor and any individual or Legal Entity + on behalf of whom a Contribution has been received by Licensor and + subsequently incorporated within the Work. + + 2. Grant of Copyright License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + copyright license to reproduce, prepare Derivative Works of, + publicly display, publicly perform, sublicense, and distribute the + Work and such Derivative Works in Source or Object form. + + 3. Grant of Patent License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + (except as stated in this section) patent license to make, have made, + use, offer to sell, sell, import, and otherwise transfer the Work, + where such license applies only to those patent claims licensable + by such Contributor that are necessarily infringed by their + Contribution(s) alone or by combination of their Contribution(s) + with the Work to which such Contribution(s) was submitted. If You + institute patent litigation against any entity (including a + cross-claim or counterclaim in a lawsuit) alleging that the Work + or a Contribution incorporated within the Work constitutes direct + or contributory patent infringement, then any patent licenses + granted to You under this License for that Work shall terminate + as of the date such litigation is filed. + + 4. Redistribution. You may reproduce and distribute copies of the + Work or Derivative Works thereof in any medium, with or without + modifications, and in Source or Object form, provided that You + meet the following conditions: + + (a) You must give any other recipients of the Work or + Derivative Works a copy of this License; and + + (b) You must cause any modified files to carry prominent notices + stating that You changed the files; and + + (c) You must retain, in the Source form of any Derivative Works + that You distribute, all copyright, patent, trademark, and + attribution notices from the Source form of the Work, + excluding those notices that do not pertain to any part of + the Derivative Works; and + + (d) If the Work includes a "NOTICE" text file as part of its + distribution, then any Derivative Works that You distribute must + include a readable copy of the attribution notices contained + within such NOTICE file, excluding those notices that do not + pertain to any part of the Derivative Works, in at least one + of the following places: within a NOTICE text file distributed + as part of the Derivative Works; within the Source form or + documentation, if provided along with the Derivative Works; or, + within a display generated by the Derivative Works, if and + wherever such third-party notices normally appear. The contents + of the NOTICE file are for informational purposes only and + do not modify the License. You may add Your own attribution + notices within Derivative Works that You distribute, alongside + or as an addendum to the NOTICE text from the Work, provided + that such additional attribution notices cannot be construed + as modifying the License. + + You may add Your own copyright statement to Your modifications and + may provide additional or different license terms and conditions + for use, reproduction, or distribution of Your modifications, or + for any such Derivative Works as a whole, provided Your use, + reproduction, and distribution of the Work otherwise complies with + the conditions stated in this License. + + 5. Submission of Contributions. Unless You explicitly state otherwise, + any Contribution intentionally submitted for inclusion in the Work + by You to the Licensor shall be under the terms and conditions of + this License, without any additional terms or conditions. + Notwithstanding the above, nothing herein shall supersede or modify + the terms of any separate license agreement you may have executed + with Licensor regarding such Contributions. + + 6. Trademarks. This License does not grant permission to use the trade + names, trademarks, service marks, or product names of the Licensor, + except as required for reasonable and customary use in describing the + origin of the Work and reproducing the content of the NOTICE file. + + 7. Disclaimer of Warranty. Unless required by applicable law or + agreed to in writing, Licensor provides the Work (and each + Contributor provides its Contributions) on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or + implied, including, without limitation, any warranties or conditions + of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A + PARTICULAR PURPOSE. You are solely responsible for determining the + appropriateness of using or redistributing the Work and assume any + risks associated with Your exercise of permissions under this License. + + 8. Limitation of Liability. In no event and under no legal theory, + whether in tort (including negligence), contract, or otherwise, + unless required by applicable law (such as deliberate and grossly + negligent acts) or agreed to in writing, shall any Contributor be + liable to You for damages, including any direct, indirect, special, + incidental, or consequential damages of any character arising as a + result of this License or out of the use or inability to use the + Work (including but not limited to damages for loss of goodwill, + work stoppage, computer failure or malfunction, or any and all + other commercial damages or losses), even if such Contributor + has been advised of the possibility of such damages. + + 9. Accepting Warranty or Additional Liability. While redistributing + the Work or Derivative Works thereof, You may choose to offer, + and charge a fee for, acceptance of support, warranty, indemnity, + or other liability obligations and/or rights consistent with this + License. However, in accepting such obligations, You may act only + on Your own behalf and on Your sole responsibility, not on behalf + of any other Contributor, and only if You agree to indemnify, + defend, and hold each Contributor harmless for any liability + incurred by, or claims asserted against, such Contributor by reason + of your accepting any such warranty or additional liability. + + END OF TERMS AND CONDITIONS + + APPENDIX: How to apply the Apache License to your work. + + To apply the Apache License to your work, attach the following + boilerplate notice, with the fields enclosed by brackets "[]" + replaced with your own identifying information. (Don't include + the brackets!) The text should be enclosed in the appropriate + comment syntax for the file format. We also recommend that a + file or class name and description of purpose be included on the + same "printed page" as the copyright notice for easier + identification within third-party archives. + + Copyright [yyyy] [name of copyright owner] + + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. diff --git a/research/relay1256-sharp-scalar-range/NOTICE b/research/relay1256-sharp-scalar-range/NOTICE new file mode 100644 index 0000000000..7e1c0160c7 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/NOTICE @@ -0,0 +1,3 @@ +Apache License 2.0. Prepared by Ali Romman with substantial OpenAI Codex assistance. + +This package adds a sharp normalized scalar-range proof and checker for the fixed role-1256 graph. Clear1nsight supplied that graph in PR #7704; SYLforge supplied its nine-swap source word in PR #7700. Preserve CREDITS.md, LICENSE and the original notices in credits/ when redistributing the graph. The retained sources credit DreamingOfClouds, Ali Romman, Rohan Arun, Jerry Chen (CubeJerry), Douglas Colkitt and the earlier community lineage. No complete multiplication theorem or selected-frontier change is claimed. diff --git a/research/relay1256-sharp-scalar-range/PROOF.md b/research/relay1256-sharp-scalar-range/PROOF.md new file mode 100644 index 0000000000..a591730f1c --- /dev/null +++ b/research/relay1256-sharp-scalar-range/PROOF.md @@ -0,0 +1,144 @@ +# Sharp normalized scalar ranges + +## Contract and result + +Fix the candidate graph from Clear1nsight's +[PR #7704](https://github.com/CrocSwap/integer-mult-bounds/pull/7704), +at `e29a5ae8591243c1178067727f12ce4abefa0b93`. +Its decompressed SHA-256 is +`8e7b755bd2d2d387bc3d3adfe48100985adf74775ca5acbb143170a3d16b5ccf`. +Keep its 1501 independent entrances, all scalar coefficients, the row +schedule and the increasing donor order within each row. A paid move is +the identity in these normalized scalar coordinates. This contract does +not identify a physical chart map with the identity. + +For independent real entrances with absolute value at most one, the exact +largest absolute values are as follows. A term number starts at one. + +| Quantity | Sharp maximum | Row | Term | Extra signed bits | +|---|---:|---:|---:|---:| +| Completed row or entrance | `2140811/96` | 101894 | 4 | 15 | +| Ordered partial sum | `53895757/256` | 156140 | 11 | 18 | +| Separate rational product | `4340699/24` | 156140 | 3 | 18 | +| Integer numerator before division | `36262265/32` | 100773 | 9 | 21 | + +The numerator uses the coefficient in reduced form `a/b`, with `b>0`. +Its value is `a` times the donor, before division by `b`. The last row +of the table comes from donor 93930 and coefficient `7345/16`. + +The source's sum of separate absolute product bounds is `13819451/16`. +This is the source receipt's `27984388275/16` divided by its odd scale +2025. The exact prefix bound is smaller by the factor +`221111216/53895757`. This is a tighter bound, not an error in the source. +The large prefix occurs in terminal row 156140. Its completed value is +exactly entrance 1372, so a bound on completed values alone misses the +temporary growth. + +For integer entrances that are multiples of 921600 and fit a signed +W-bit word, `W+21` signed data bits suffice for the stated normalized +scalar arithmetic, for all `W>=2`. A 32-bit entrance fixture needs all +21 extra bits. This does not replace the source's physical `W+26` +service contract: its W measures a different domain of chart-valued +donors and inherited root states. + +## Linear forms give both bounds and witnesses + +Let `c(v)` be the rational coefficient vector of any intermediate v in +the independent entrances. Each entrance has its unit vector. If the +literal terms of a row are `q_j v_j`, then its prefix after t terms has +coefficient vector + +```text +c(prefix_t) = sum_(j=1)^t q_j c(v_j). +c(product_j) = q_j c(v_j). +c(numerator_j) = numerator(q_j) c(v_j). +``` + +These identities follow by induction in the source schedule. They do +not assume independent donors within a row. All appearances of the same +entrance must first be combined. The stored row uses the complete sum. + +For a real vector x with every `|x_i|<=1`, the triangle inequality gives +`|c dot x| <= sum_i |c_i|`. Set `x_i=sign(c_i)` at each nonzero +coefficient, and set all other entries to one. This input attains the +bound. Thus the coefficient l1 norm is the exact range of each linear +form. The largest such norm over a finite set of forms is also attained. +This proves the all-input statement; the finite computation only has to +enumerate the fixed source forms correctly. + +The checker represents `921600*c(v)` by integers. Every basis product +divides exactly. It updates the prefix norm by `|old+product|-|old|` +at each affected entrance. After each complete donor term, it checks +that prefix against the maximum. It also checks every separate product, +every numerator, all completed rows and all initial unit vectors. +The enumeration covers 7644 rows, 172414 ordered prefixes and 9044092 +nonzero basis products. Source hashes, schedule completeness, unique +donor order, causality and donor charts are checked before this pass. + +The receipt gives the negative entrance columns of each attaining input. +All other signs are one. Four complete scalar replays use these inputs +with rational arithmetic, without coefficient vectors or the norm update. +Each attains its listed maximum and stays within all four bounds. +Four further complete scalar replays use the integer fixtures below. + +## Integer range and sharp bit counts + +If all entrance integers are multiples of G=921600, the integer basis +pass proves that every product and stored value is integral. Prefixes +are sums of those integral products. No rounding occurs. Let C be one +of the four exact unit-input bounds. Signed W-bit entrances have absolute +value at most `2^(W-1)`, so the corresponding intermediates have absolute +value at most `C*2^(W-1)`. + +For the four rows of the table, respectively, + +```text +2^14 < 2140811/96 < 2^15 +2^17 < 53895757/256 < 2^18 +2^17 < 4340699/24 < 2^18 +2^20 < 36262265/32 < 2^21. +``` + +The strict upper inequalities imply the stated signed widths. To check +that one fewer bit does not work uniformly, set W=32 and set every +entrance to its attaining sign times +`M=921600*floor((2^31-1)/921600)=2147328000`. Both signs fit W bits. +Every intermediate remains integral. The resulting target values are: + +| Quantity | Attained integer | Largest positive value with one fewer bit | +|---|---:|---:| +| Completed row | 47885660448000 | 35184372088831 | +| Ordered partial sum | 452077609716000 | 281474976710655 | +| Separate product | 388371020928000 | 281474976710655 | +| Numerator | 2433343030560000 | 2251799813685247 | + +These are reachable values of the actual graph, rather than independent +choices for donors that may share entrances. In particular, checking only +product and prefix bounds would suggest 18 extra bits and miss the +numerator requirement. The 20-bit numerator allowance also fails. + +A signed shift-and-add multiplication can form `a*v` with no absolute +partial product larger than `|a*v|`: operate on the absolute values, +then apply the sign. In binary long division by positive b, every +quotient prefix is at most the absolute numerator. Before each possible +subtraction, the remainder is less than `2b`. The source has `b<=384`, +so these remainders are less than 768. These values fit `W+21` for +`W>=2`. Products, sum operands and sums have the bounds above. This +gives an arithmetic data-width construction. It does not price a tape +implementation, descriptor service, child workspace or bank traversal. + +## Verification limits + +The sharp values apply to one fixed normalized scalar graph and its +literal donor order. Different ordering, coefficients, graph or entrance +contract require a new check. The full source endpoint theorem is not +reproved here; the range theorem is about the supplied graph itself. +The terminal identity at the prefix peak is checked as a control. + +Physical inputs on different charts involve the operators +`T_current * inverse(T_input)`. Their norms, paid-child internal states, +root scaling and precision recurrence remain separate. The present +scalar bounds cannot be substituted for those obligations without a +proof. No paid profile, caller invoice, complete kappa, selected record +or inherited assumption changes. No independent human or formal-kernel +review is claimed. diff --git a/research/relay1256-sharp-scalar-range/README.md b/research/relay1256-sharp-scalar-range/README.md new file mode 100644 index 0000000000..3bafc076e3 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/README.md @@ -0,0 +1,78 @@ +# Sharp ranges for the role-1256 scalar graph + +This package proves the exact largest normalized scalar values in the +fixed graph from [PR #7704](https://github.com/CrocSwap/integer-mult-bounds/pull/7704). +The sharp ordered-prefix bound is `53895757/256`, versus the source's +separate-product triangle bound `13819451/16`. The graph and donor order +stay fixed. No new kappa or selected-frontier change follows. + +Read PROOF.md for the contract and RESULT.json for all attaining inputs. +The proof covers arbitrary real entrance values in the unit cube. The +finite check covers all 7644 rows, 172414 ordered prefixes and 9044092 +nonzero basis products. Eight complete scalar replays verify the four +attaining corners both over the rationals and on an integer lattice. + +The sharp completed-row, product and predivision-numerator bounds are +`2140811/96`, `4340699/24` and `36262265/32`. For signed W-bit integer +entrances that are multiples of 921600, the normalized arithmetic needs +at most `W+21` signed data bits. A reachable 32-bit entrance fixture +fails with `W+20`. The separate sharp growth counts for completed rows, +prefixes, products and numerators are 15, 18, 18 and 21 bits. + +The source's physical `W+26` service uses a different W contract. Physical +chart norms, paid-child internal states, root scaling, tape costs and +compiler substitution remain open here. The five selected assumptions +remain. This is an author-checked source-bound arithmetic result, not a +full endpoint audit, supplier admission or formal multiplication proof. + +## Reproduction + +Use CPython 3.12 and the standard library. From the repository root: + +```sh +python3 -B scripts/research_package.py --package research/relay1256-sharp-scalar-range +``` + +For a standalone copy, run both commands: + +```sh +python3 -I -W error -B research/relay1256-sharp-scalar-range/verify.py +python3 -I -W error -B -O research/relay1256-sharp-scalar-range/verify.py +``` + +Each prints: + +```text +PASS_RELAY1256_SHARP_SCALAR_RANGE +7644 rows; 172414 prefixes; 9044092 basis products; 8 scalar oracle runs +Extra signed bits: snapshots 15; prefixes 18; products 18; numerators 21 +``` + +The checker reads committed bytes and never rewrites the receipt. It +checks source hashes, chronology, donor charts, exact integer divisions, +all ranges, the source triangle bound, four attaining corners, four +scaled integer corners and four one-bit-smaller failure witnesses. It +also checks that the row with the largest prefix ends as entrance 1372. +This detects a bound that considers completed rows alone. + +SOURCES.json binds the main-branch catalogue and selected record. The +unmerged graph is copied from an explicit public commit and checked by +both compressed and decompressed hashes. CREDITS.md records that source; +it is not an inferred ancestor of main. Its full predecessor suites, +selected cold replay, physical application, native operators, Lean, +Docker and independent human/formal-kernel review were not run here. + +## Open next questions + +1. Can a bound on `T_current * inverse(T_input)` promote the exact scalar + cancellation to a physical chart-valued range bound? Start with the + shared factor for each entrance column. Child interiors and the + common-root precision recurrence must still be covered. +2. Can a fully charged physical service use the smaller normalized + range? First specify its input word contract and division temporaries. + The present arithmetic proof supplies no tape or compiler cost gain. + +Prepared by Ali Romman with substantial OpenAI Codex assistance under +Apache-2.0. Clear1nsight supplied the graph and source amplitude bound; +SYLforge supplied the word. CREDITS.md and credits/ preserve their +contributions and the bank, caller and framework lineage. diff --git a/research/relay1256-sharp-scalar-range/RESULT.json b/research/relay1256-sharp-scalar-range/RESULT.json new file mode 100644 index 0000000000..18af91c46f --- /dev/null +++ b/research/relay1256-sharp-scalar-range/RESULT.json @@ -0,0 +1,832 @@ +{ + "coverage": { + "completed_rows_sha256": "20e47815fca6c1d098a5b618d8897275aabfd20cef98aff67dcfb473d235f903", + "nonzero_basis_products": 9044092, + "ordered_prefixes": 172414, + "peak_prefix_row_final_value": "input 1372", + "rows": 7644, + "separate_product_triangle_maximum": "13819451/16", + "triangle_peak_row": 156140 + }, + "integer_entrance_multiple": 921600, + "long_division_remainder_strict_upper": 768, + "maximum_coefficient_denominator": 384, + "one_bit_smaller_controls": 4, + "scalar_oracle_complete_runs": 8, + "scope": "Fixed normalized scalar graph; no physical chart or complete kappa claim.", + "sharp_ranges": { + "numerator": { + "all_other_input_signs": 1, + "attaining_integer_value": 2433343030560000, + "coefficient_vector_sha256": "597bfe76c9639cb7b9b97c038988382c16e64b409717e7d45b07597072f8bb57", + "input_integer_magnitude": 2147328000, + "input_signed_width": 32, + "maximum_for_unit_inputs": "36262265/32", + "negative_input_columns": [ + 292, + 360, + 471, + 484, + 485, + 512, + 573, + 933, + 976, + 979, + 1282 + ], + "one_based_term": 9, + "one_bit_smaller_positive_limit": 2251799813685247, + "row": 100773, + "source_term": [ + 93930, + "7345/16" + ], + "sufficient_uniform_extra_signed_bits": 21, + "support": 20 + }, + "prefix": { + "all_other_input_signs": 1, + "attaining_integer_value": 452077609716000, + "coefficient_vector_sha256": "a35a1bb7eb7362aeb5bd8f4a0f5413e4974c0e45eecb40e3a268573f0875f783", + "input_integer_magnitude": 2147328000, + "input_signed_width": 32, + "maximum_for_unit_inputs": "53895757/256", + "negative_input_columns": [ + 2, + 3, + 5, + 6, + 8, + 10, + 12, + 13, + 15, + 16, + 18, + 20, + 22, + 30, + 35, + 36, + 39, + 40, + 41, + 43, + 47, + 48, + 49, + 51, + 52, + 54, + 58, + 59, + 63, + 64, + 68, + 70, + 73, + 78, + 80, + 82, + 83, + 85, + 90, + 94, + 96, + 101, + 103, + 104, + 105, + 107, + 108, + 111, + 126, + 131, + 172, + 173, + 229, + 233, + 239, + 243, + 247, + 248, + 249, + 252, + 253, + 255, + 260, + 265, + 272, + 281, + 284, + 288, + 292, + 294, + 300, + 306, + 311, + 312, + 313, + 316, + 331, + 334, + 336, + 344, + 350, + 354, + 360, + 363, + 364, + 365, + 366, + 367, + 370, + 372, + 376, + 382, + 383, + 388, + 398, + 404, + 409, + 411, + 414, + 416, + 418, + 421, + 428, + 448, + 461, + 469, + 471, + 479, + 484, + 485, + 490, + 499, + 510, + 512, + 518, + 525, + 527, + 528, + 530, + 536, + 537, + 538, + 539, + 543, + 544, + 545, + 547, + 550, + 551, + 558, + 561, + 562, + 564, + 566, + 568, + 569, + 587, + 591, + 599, + 608, + 618, + 619, + 621, + 624, + 635, + 643, + 655, + 659, + 664, + 680, + 697, + 704, + 707, + 711, + 732, + 752, + 760, + 775, + 802, + 830, + 929, + 990, + 991, + 1017, + 1022, + 1027, + 1033, + 1042, + 1043, + 1048, + 1060, + 1062, + 1063, + 1065, + 1068, + 1073, + 1074, + 1077, + 1078, + 1086, + 1094, + 1095, + 1104, + 1105, + 1109, + 1110, + 1112, + 1118, + 1121, + 1122, + 1126, + 1127, + 1128, + 1139, + 1141, + 1142, + 1143, + 1144, + 1160, + 1161, + 1164, + 1170, + 1172, + 1173, + 1175, + 1176, + 1177, + 1178, + 1180, + 1181, + 1184, + 1193, + 1194, + 1201, + 1203, + 1204, + 1205, + 1206, + 1207, + 1210, + 1216, + 1221, + 1227, + 1230, + 1233, + 1234, + 1235, + 1242, + 1247, + 1248, + 1249, + 1250, + 1253, + 1255, + 1259, + 1260, + 1264, + 1273, + 1274, + 1287, + 1306, + 1311, + 1318, + 1349, + 1375, + 1472 + ], + "one_based_term": 11, + "one_bit_smaller_positive_limit": 281474976710655, + "row": 156140, + "source_term": [ + 103886, + "4" + ], + "sufficient_uniform_extra_signed_bits": 18, + "support": 448 + }, + "product": { + "all_other_input_signs": 1, + "attaining_integer_value": 388371020928000, + "coefficient_vector_sha256": "971ecfb5058a3a556844b2c73933edd907675f873ad5520e19b46e802f3c3972", + "input_integer_magnitude": 2147328000, + "input_signed_width": 32, + "maximum_for_unit_inputs": "4340699/24", + "negative_input_columns": [ + 3, + 5, + 6, + 9, + 10, + 15, + 18, + 22, + 25, + 30, + 36, + 40, + 41, + 47, + 49, + 52, + 54, + 59, + 63, + 64, + 68, + 70, + 73, + 79, + 82, + 83, + 89, + 90, + 96, + 101, + 105, + 108, + 111, + 131, + 172, + 173, + 233, + 238, + 239, + 243, + 247, + 249, + 252, + 253, + 255, + 265, + 267, + 272, + 274, + 281, + 284, + 288, + 292, + 294, + 300, + 306, + 311, + 313, + 315, + 316, + 317, + 331, + 334, + 336, + 344, + 354, + 360, + 363, + 364, + 365, + 366, + 370, + 372, + 376, + 382, + 383, + 388, + 398, + 399, + 404, + 408, + 409, + 414, + 418, + 428, + 448, + 469, + 471, + 479, + 484, + 485, + 490, + 499, + 510, + 512, + 518, + 527, + 528, + 530, + 536, + 537, + 538, + 539, + 543, + 544, + 545, + 547, + 550, + 551, + 558, + 561, + 562, + 564, + 565, + 566, + 567, + 568, + 569, + 578, + 587, + 591, + 599, + 608, + 611, + 618, + 619, + 621, + 624, + 635, + 643, + 655, + 659, + 664, + 680, + 697, + 704, + 707, + 711, + 732, + 752, + 760, + 775, + 802, + 830, + 929, + 990, + 991, + 1017, + 1022, + 1027, + 1033, + 1042, + 1043, + 1048, + 1060, + 1062, + 1063, + 1065, + 1068, + 1073, + 1074, + 1077, + 1078, + 1084, + 1086, + 1094, + 1095, + 1105, + 1110, + 1112, + 1118, + 1121, + 1122, + 1126, + 1139, + 1141, + 1142, + 1143, + 1144, + 1160, + 1161, + 1164, + 1170, + 1172, + 1173, + 1175, + 1176, + 1177, + 1178, + 1193, + 1194, + 1201, + 1204, + 1205, + 1206, + 1210, + 1216, + 1221, + 1227, + 1230, + 1233, + 1234, + 1235, + 1242, + 1247, + 1248, + 1249, + 1250, + 1253, + 1256, + 1259, + 1260, + 1272, + 1273, + 1274, + 1287, + 1306, + 1311, + 1318, + 1375, + 1472 + ], + "one_based_term": 3, + "one_bit_smaller_positive_limit": 281474976710655, + "row": 156140, + "source_term": [ + 103582, + "-44" + ], + "sufficient_uniform_extra_signed_bits": 18, + "support": 394 + }, + "snapshot": { + "all_other_input_signs": 1, + "attaining_integer_value": 47885660448000, + "coefficient_vector_sha256": "56aa654087cb1d904b1c947541f2e2390d5e2ee45dcca2456a072dbb86d5b400", + "input_integer_magnitude": 2147328000, + "input_signed_width": 32, + "maximum_for_unit_inputs": "2140811/96", + "negative_input_columns": [ + 3, + 6, + 13, + 15, + 36, + 41, + 43, + 44, + 47, + 52, + 68, + 73, + 75, + 83, + 96, + 99, + 101, + 105, + 129, + 131, + 142, + 150, + 152, + 172, + 173, + 230, + 238, + 239, + 243, + 247, + 248, + 249, + 252, + 255, + 264, + 265, + 267, + 272, + 280, + 281, + 283, + 288, + 294, + 300, + 302, + 306, + 311, + 313, + 315, + 316, + 336, + 352, + 356, + 360, + 363, + 365, + 366, + 372, + 383, + 388, + 398, + 399, + 408, + 409, + 414, + 428, + 434, + 441, + 448, + 458, + 461, + 469, + 472, + 473, + 479, + 482, + 483, + 484, + 491, + 494, + 497, + 499, + 502, + 504, + 511, + 512, + 517, + 518, + 527, + 528, + 530, + 536, + 538, + 539, + 543, + 544, + 545, + 550, + 551, + 560, + 561, + 562, + 563, + 564, + 565, + 566, + 567, + 568, + 569, + 570, + 571, + 576, + 577, + 578, + 587, + 599, + 600, + 608, + 619, + 624, + 635, + 640, + 643, + 664, + 681, + 688, + 697, + 704, + 710, + 711, + 716, + 732, + 752, + 760, + 768, + 775, + 810, + 830, + 882, + 883, + 888, + 894, + 895, + 896, + 897, + 898, + 910, + 918, + 919, + 920, + 921, + 928, + 929, + 930, + 931, + 932, + 933, + 934, + 949, + 950, + 951, + 954, + 958, + 959, + 960, + 968, + 975, + 976, + 977, + 978, + 980, + 981, + 982, + 988, + 991, + 1009, + 1017, + 1022, + 1027, + 1032, + 1033, + 1034, + 1035, + 1042, + 1043, + 1048, + 1055, + 1058, + 1060, + 1062, + 1063, + 1065, + 1072, + 1073, + 1074, + 1084, + 1104, + 1105, + 1106, + 1112, + 1113, + 1122, + 1126, + 1139, + 1141, + 1142, + 1143, + 1144, + 1160, + 1161, + 1164, + 1170, + 1172, + 1173, + 1174, + 1175, + 1176, + 1177, + 1178, + 1186, + 1188, + 1190, + 1198, + 1204, + 1205, + 1206, + 1210, + 1216, + 1227, + 1233, + 1234, + 1235, + 1236, + 1240, + 1243, + 1247, + 1249, + 1250, + 1253, + 1256, + 1257, + 1258, + 1272, + 1273, + 1274, + 1287, + 1291, + 1306, + 1311, + 1328, + 1349, + 1449, + 1469, + 1470, + 1471, + 1472 + ], + "one_based_term": 4, + "one_bit_smaller_positive_limit": 35184372088831, + "row": 101894, + "source_term": [ + 101892, + "1" + ], + "sufficient_uniform_extra_signed_bits": 15, + "support": 492 + } + }, + "source_graph_sha256": "8e7b755bd2d2d387bc3d3adfe48100985adf74775ca5acbb143170a3d16b5ccf", + "source_triangle_to_sharp_prefix_ratio": "221111216/53895757", + "status": "PASS_RELAY1256_SHARP_SCALAR_RANGE" +} diff --git a/research/relay1256-sharp-scalar-range/SOURCES.json b/research/relay1256-sharp-scalar-range/SOURCES.json new file mode 100644 index 0000000000..5335fa97d9 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/SOURCES.json @@ -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" + } + ] +} diff --git a/research/relay1256-sharp-scalar-range/TASK.json b/research/relay1256-sharp-scalar-range/TASK.json new file mode 100644 index 0000000000..ed5fbbdd9a --- /dev/null +++ b/research/relay1256-sharp-scalar-range/TASK.json @@ -0,0 +1,89 @@ +{ + "schema_version": 1, + "task_id": "uniform-transfer-audit", + "priority_revision": 8, + "base_commit": "e54523517d0f03e58352e647c403a5af3e525a57", + "original_card": { + "id": "uniform-transfer-audit", + "question": "Can one of the selected record's five inherited assumptions be proved at its actual caller?", + "starting_points": [ + "docs/research/current-priorities.md", + "docs/research/research-family-status.md", + "certificates/selected-result.json", + "docs/research/pr-intake-evidence/pr5599-completion/PROOF-REVIEW.md" + ], + "methods": [ + "Construct and independently check one source-bound instance of the stated first experiment.", + "Seek an exact counterexample or target-specific complete-budget obstruction to that same proposed change." + ], + "acceptance_gate": "Identify the proposition and what would change downstream; do not equate a finite replay with a universal proof.", + "first_step": { + "command": [ + "python3", + "-B", + "scripts/research_selected_baseline.py" + ], + "expected": "PASS_RESEARCH_BASELINE_BINDINGS_ONLY", + "inspect": "Read revision 8, the selected review and live exact-source handoffs. Then follow this card's bounded first_experiment.", + "scope": "Source-bound orientation only, not a supplier replay or theorem proof." + }, + "first_experiment": "Choose one assumption ID (native_all_columns, owned_erasure, compiler, precision_tape, analytic_recovery), one quantified implication and exact dependency closure. Check current handoffs, including #5625 row reentry and #5707 root-prefix lowering, without treating their partial scope as the whole theorem. Prove or falsify that implication with a caller-bound argument; stop at the chosen obligation. Finite tests or root digits do not discharge an all-size statement.", + "bottleneck_link": "Judge the actual current supplier contract and complete signed target/paid bill; frozen-profile ceilings require an applicability proof.", + "allowed_variations": [ + "One exact source head, helper/cut region or quantified proposition and changed hypothesis; check live handoffs first. Different seeds are not different scopes." + ], + "scope_change_triggers": [ + "Changing source head, helper/cut domain, operation alphabet or caller boundary without recording the new scope.", + "Treating an archived proposal or local numerical pass as complete supplier admission." + ], + "overlap_terms": [ + "native_all_columns", + "owned_erasure", + "compiler", + "precision_tape", + "analytic_recovery" + ], + "existing_checks": [ + { + "scope": "PR #4100: proposed unconditional construction and finite checks, without independent acceptance.", + "evidence": "research/unconditional-batched-crt/README.md", + "remaining": "Independent proof audit of one caller-bound all-size implication; all other premises remain explicit.", + "status": "author-checked", + "source_commit": "337d04939571a637aeab54a0f58f54fde811663a", + "selected_record_sha256": "bb2eaa58647b353f88d0aec2eafb9d30df1be53589231b23e12ceef3e4bad129" + }, + { + "scope": "Selected PR5599: complete helper-subspace construction and independent caller replay; five framework assumptions remain.", + "evidence": "docs/research/pr-intake-evidence/pr5599-completion/PROOF-REVIEW.md", + "remaining": "The next experiment must identify uncovered source-specific work. Successor archive intake does not transfer this admission.", + "status": "reviewed", + "source_commit": "09dd0965711ca7eeb169057a420abab1c9052137", + "selected_record_sha256": "2a0220e866417774606e0f991e41b9c1d26b1f43b97857238cf478dbc7d330b5" + } + ] + }, + "scope_status": "changed", + "actual_scope": "Sharp normalized scalar ranges and reachable integer overflow witnesses for the fixed PR7704 role-1256 graph.", + "scope_change_reason": "The current supplier review leaves source-specific precision obligations. This finite arithmetic audit retains all graph dependencies and separates chart/root obligations.", + "bottleneck_link": "A range proof must cover ordered partial sums and integer numerators, not only completed rows. This is a finite precision_tape sub-obligation.", + "budget_minutes": 60, + "stop_rule": "Stop at sharp scalar ranges and complete attaining-input replays; do not change the graph, physical service, invoice or selected frontier.", + "outcome": "completed", + "evidence": "PROOF.md; RESULT.json; all-row basis replay; eight complete scalar oracles; four one-bit-smaller failure witnesses.", + "author": "Ali Romman", + "ai_assistance": "Substantial OpenAI Codex assistance. The graph and notices are copied and attributed; the source verifier is not imported.", + "open_questions": [ + { + "question": "Can the scalar cancellation be promoted to a chart-valued physical range bound?", + "why_it_matters": "It could tighten a real precision obligation only after chart norms and child states are covered.", + "first_step": "Bound the common operator factor for each independent entrance column.", + "blocked_by": "Paid-child interiors and the common-root recurrence remain separate." + }, + { + "question": "Can a fully charged service consume the smaller normalized range?", + "why_it_matters": "A local bit bound alone does not reduce the complete ordinary charge.", + "first_step": "Specify the physical entrance contract, numerator temporaries and tape operations.", + "blocked_by": "The current package proves no physical compiler substitution." + } + ] +} diff --git a/research/relay1256-sharp-scalar-range/VERIFY.json b/research/relay1256-sharp-scalar-range/VERIFY.json new file mode 100644 index 0000000000..960b478384 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/VERIFY.json @@ -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": 120, + "checks": "Authenticate the fixed graph; enumerate every scalar range; check all attaining rational and integer inputs and four one-bit-smaller failure witnesses." + }, + { + "argv": [ + "{python}", + "-I", + "-W", + "error", + "-B", + "-O", + "{package}/verify.py" + ], + "timeout_seconds": 120, + "checks": "Authenticate the fixed graph; enumerate every scalar range; check all attaining rational and integer inputs and four one-bit-smaller failure witnesses." + } + ] +} diff --git a/research/relay1256-sharp-scalar-range/credits/7658-NOTICE b/research/relay1256-sharp-scalar-range/credits/7658-NOTICE new file mode 100644 index 0000000000..2baa2a719e --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/7658-NOTICE @@ -0,0 +1,21 @@ +Apache-2.0. DreamingOfClouds with Claude Opus 5.5 (Anthropic), via Claude Code. + +The base word (data/base-7641-word.json.gz, unmodified bytes, sha256 de43ed04f58133eaef8f8d12a9dae1e6e8aa7f9721044200cc839673d86fd7c1) +is the word of our #7641 (20 swap-then-delete composites on SYLforge's #7637), vendored as data only. #7641's own base is SYLforge's +#7637 candidate word, whose notices read: + + "Apache-2.0. Local H5B + E1/E4 combination experiment, prepared with OpenAI assistance. + All earlier contributors and notices in hydrogen-bounded-verification-20261117/NOTICE remain credited. In particular: Ali Romman + (#7605), SYLforge, Clear1nsight, DreamingOfClouds, and all upstream contributors listed there." + +#7637's recorded technique credits: DreamingOfClouds #7630 (E1 rebirth/canceller, E4 existing-co-presence reroute), Clear1nsight +#7631/#7633 (efficient E4 integration, E1 prefix/tail settlement), and SYLforge's H5B literal-frame-restored archive. Its lineage +also carries Clear1nsight's #7626/#7314/#7274, Rohan Arun's #7277 and signed #7113 caller, and the contributors named in the +retained sources. #7641, #7637 and their code are not run by this package; only their words are read as data. + +This package adds five edits found by a large-neighbourhood search (one SWAP-NULL relabel of two helpers with a stay deletion, one main +stop removal, three helper stay deletions with re-placed contacts) and one hub re-settle. The helper-stay deletion with re-charted +contacts (E3) is of the type introduced by Clear1nsight's #7612 and ToikY's #7643. check.py is our own code: its replay, crossing and +column-test sections come from our #7630-on-#7623, swap-composite and #7641 packages; its column tests follow the description of the +affected-pair tests in #7620's README and native runner. The CLEAR certificate check implements the acceptance rule of #7620's +verify_certificates (read, not run). No other author's code was run; their words are read as data. diff --git a/research/relay1256-sharp-scalar-range/credits/7700-CITATION.cff b/research/relay1256-sharp-scalar-range/credits/7700-CITATION.cff new file mode 100644 index 0000000000..d80f6b8660 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/7700-CITATION.cff @@ -0,0 +1,55 @@ +cff-version: 1.2.0 +message: >- + Please cite the selected composition and its constituent contributions, + this project and the upstream manuscript. The selected checkpoint is PR4470 on the cumulative productive-helper framework; consult certificates/selected-result.json for publication status. + Project-level authorship identifies the maintainer, not sole authorship of + the community result. See CONTRIBUTORS.md and the frozen source notices. +title: "A sharper exponent for integer multiplication" +authors: + - family-names: Colkitt + given-names: Douglas +license: Apache-2.0 +type: software +version: "community-joint4470-2026-10-22" +date-released: 2026-10-22 +repository-code: "https://github.com/CrocSwap/integer-mult-bounds" +url: "https://github.com/CrocSwap/integer-mult-bounds" +abstract: >- + Community research on integer multiplication. The selected maintainer-reviewed + conditional construction has kappa = 60403/20000000 = 0.00302015. + Cinna Davis's PR4470 joint helper frames and query-thinned controller build on + her preceding constructions, zxv's native/productive-helper and compiler work, + Jacob Alan Sussman's mixed-target recursion, and the cumulative community + framework. Exact finite replays and independent audits retain explicit + all-column/native, owned-erasure, compiler, precision/tape, analytic, prime + and recovery assumptions. This is not a complete formal multiplication proof. +preferred-citation: + type: software + title: "A sharper exponent for integer multiplication: community research repository" + authors: + - family-names: Colkitt + given-names: Douglas + year: 2026 + url: "https://github.com/CrocSwap/integer-mult-bounds" + notes: >- + Maintainer citation only, not sole authorship of the mathematical result. + Credit Cinna Davis (PR4470/4447/4441/4440/4416), zxv (PR4359/4433), + Jacob Alan Sussman (PR4268), independent auditors Mihir Srivastava and + Ryanson Jonathan, Mark/Matarener for the commutation question, and the full + geometry, binary, native, precision and outer lineage, including Rohan Arun, + Jeremy Kothe, Chafik Boukhalfa and the contributors listed in CONTRIBUTORS.md + and docs/research/community-joint4470-contributors.md. Cite constituent PRs + where their results are used; preserve their licenses and assistance disclosures. +keywords: + - integer multiplication + - computational complexity + - exact arithmetic + - parameter optimization + - reproducible research +references: + - type: unpublished + title: "Integer multiplication below n log n" + authors: + - name: OpenAI + year: 2026 + url: "https://github.com/openai/math/tree/adc7f1241b42e322a6451854ab7e4b4c146bf78a/preprints/Integer-multiplication-below-n-log-n-September-23-2026" diff --git a/research/relay1256-sharp-scalar-range/credits/7700-NOTICE b/research/relay1256-sharp-scalar-range/credits/7700-NOTICE new file mode 100644 index 0000000000..59d8e0cf31 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/7700-NOTICE @@ -0,0 +1,415 @@ +A sharper exponent for integer multiplication +Copyright 2026 Douglas Colkitt + +This project includes unmodified source files from OpenAI's manuscript +"Integer multiplication below n log n" and its accompanying repository: +https://github.com/openai/math +Commit: adc7f1241b42e322a6451854ab7e4b4c146bf78a + +The original manuscript identifies OpenAI as its author. Its Apache-2.0 +license is retained at upstream/LICENSE. The exact source URLs and hashes +are recorded in upstream/manifest.json. Proposed modifications are supplied +as separate files under patches/; upstream source files remain unchanged. + +The original local research note, scripts, certificates, patches, and +documentation were prepared by Douglas Colkitt with assistance from OpenAI +Codex. The community release incorporates work by the contributors identified +in the notices below; it is not solely Douglas Colkitt's work. +This project is not an official OpenAI release or an OpenAI endorsement. + +Related contributions (not imported in this checkpoint): +Zhihao Chen (jacklightChen), PR #7, supplied an earlier ternary five-subset +motif, finite-alphabet transfer and a different compressed producer: +https://github.com/CrocSwap/integer-mult-bounds/pull/7 +The present checkpoint claims no priority for that motif. Related compressed +complex-network work was submitted by eumemic in PR #3: +https://github.com/CrocSwap/integer-mult-bounds/pull/3 +The pending contributions and their stronger bounds require their own review; +this notice does not represent adoption or verification of their claims. + +Additional community work, including parallel and superseded submissions, is +acknowledged in CONTRIBUTORS.md, with pinned submission metadata in +docs/research/contribution-snapshot.json. These acknowledgements do not imply +joint authorship of the checkpoint, incorporation of pending code, or verification +of submitted mathematical claims. Original source-specific notices remain in force. + +Community release (accepted by conditional maintainer audit): +The following original contributor notices apply to the imported PR #39 chain. +The preceding non-import statements describe the preserved 2^-30 checkpoint. + +Ternary five-subset construction and common-four-point resynthesis +Copyright 2026 Zhihao Chen (jacklightChen) + +This is Zhihao Chen's first submission of this specific construction to this +repository. Research, proof development, code and verification were prepared +with substantial assistance from GPT-6 Astra (OpenAI Codex), as identified by +the contributor. This is not independent review or an OpenAI endorsement. +Future research building on this contribution should explicitly acknowledge +Zhihao Chen (jacklightChen), cite the accompanying note and pull request, and +retain the preceding contributors' attribution. This notice records research +provenance; the repository's Apache-2.0 license remains unchanged. + +Dependencies retained from eumemic's Claude-assisted PRs #3 and #5: +https://github.com/CrocSwap/integer-mult-bounds/pull/3 +https://github.com/CrocSwap/integer-mult-bounds/pull/5 +PR #5 commit: d3d370c34937503bd3fa63c83d505a05e8e50893 +These include complex_circuit.py, complex_network.py, fast_gaussian.py, +their generators, construction notes and certificates. The Gaussian input +scaling proof is corrected in this derivative; see the modification notice below. +Aligned pairing also appears in Bortlesboat's PR #2. Retained totals and bank +sharing appear in dleen's PR #4 and cheaper retained centers in eumemic's PR #6. +Their results and ideas are credited separately from this new construction. + +Bulk recursion and controlled tensor basis, 8 October 2026 +Copyright 2026 icekylinx + +The new projector batching, controlled simultaneous basis, mixed-width tape +recursions, whole-residual complex recursion, and dependency-path precision +guard in the batched-23 proof package are contributed by icekylinx, developed +with substantial OpenAI GPT-6 Astra assistance. Review corrections and +repository integration were prepared with OpenAI Codex assistance. They build on the retained +PR #7 finite producers and the preceding contributions credited above. +The new package retains the repository's Apache-2.0 license and the original +authors' attribution. It is not an official OpenAI release or endorsement. + +Modification notice, 8 October 2026 (icekylinx): the retained fast Gaussian +proof now bounds shifted approximation error by F=2^ceil(1.14 alpha^2). +The P=34p precision budget and asymptotic exponent are unchanged. New batching +checks are integrated into make verify. Existing attribution and licenses +are retained. + +Partial-swap extension, 8 October 2026 +Copyright 2026 icekylinx + +The partial-swap framing, binary retained-total triple construction, +dependency-preserving positive labels and carrier reuse, unequal (25,23,57) +common basis, improved complex batching, and exact assembly are contributed +by icekylinx with substantial OpenAI GPT-6 Astra and Codex assistance. +This extension directly inherits PR #10, commit +62691e395a0458ce089a1c7b5d89e74291e95e29. The portable producer adapters +modify the retained paired-exclusion and shared-point implementations credited +above. The original licenses, attribution, and Gaussian correction are retained. + +Endpoint-gauge extension, 8 October 2026 +Copyright 2026 icekylinx + +The complementary auxiliary endpoint gauges, (32,30,40) controlled basis, +general nondegenerate binary phase residuals, retained-center complex producer, +updated path guard and exact assembly are contributed by icekylinx with +substantial OpenAI GPT-6 Astra and Codex assistance. This extension inherits +f2ab41aebad47861caf6316282c1793e5513845e and its PR #10 ancestry. New portable +adapters retain the paired-exclusion and shared-point contributions credited +above. Original licenses, notices and the Gaussian correction are retained. + +Structured-basis and semantic-bulk extension, 8 October 2026 +Copyright 2026 icekylinx + +The selected (32,30,36) bit construction, A1/A5 block refinements, fixed +middle-axis I+J profiles with bounded-minor CRT certificates, mixed dyadic +complex centers, complete data-front batching and circuit-specific assembly +are contributed by icekylinx with substantial OpenAI GPT-6 Astra and Codex +assistance. The immediate parent is ed90fd940279c336ebc45968631bdebfb087b505. + +The semantic/bulk compatibility argument, literal scalar charge, product row +stock and parameterization adapt Zhihao Chen (jacklightChen)'s PR #23, +commit 661ebabc1076c9f525d7ce4967c876b203fa9827. Translated complex endpoints +are adopted from the same contributor's PR #21, +commit 5ba6cf0bfb68f2be8d15610e7972207c50254d6a, following eumemic's source-frame +work and the earlier nested-basis contributions. + +Completed-child semantic precision, arbitrary-coordinate routing, phase-cell +Gaussian inversion and bulk tape transfer are credited to the RaD project +(hipotures), commit 45d9b60355872041f9275f77c057514def0d45bc, exposed through +PR #20 at 4f8d6c8272b5ff307a0da51df545ec3cd96a8b6e. Their selected proof +sources retain their separate CC0 license and original NOTICE under +references/semantic-bulk/rad20. The PR #23 source retains its recorded OpenAI +Codex assistance; RaD retains its recorded Codex and GPT-6.1-Sol Ultra +assistance. These are source attributions, not claims about this session's +runtime model. Related contributions by Rohan Arun and Aurel Prosz remain +acknowledged in the imported notices. No exclusive priority or additional +license condition is asserted. + +Copied retained-center extension, 8 October 2026 +Copyright 2026 icekylinx + +The copied retained-center schedule in both orientations, its complex-phase +transfer, the Wm-N+L rank budget, and the selected (25,23)/(28,28) composition +are contributed by icekylinx with substantial OpenAI GPT-6 Astra and Codex +assistance. This extension inherits PR #32 at +0ef3aeb61f55cc0b321ce6a0ef00acee25cefe52 and the preceding PR #10/#18/#24 work. + +The two-stage topology and paid copied-stream endpoint correction are due +to Aurel Prosz (Paureel), c82d09eb4781b68435e3fa3eb6beea72fa900fab, +through Zhihao Chen (jacklightChen)'s PR #29 unequal-axis composition, +9d963275075fa98f1da821e27757b238dafd6b3c. That source also credits Swapnil +Jain's two-stage development at ae405eb474d1486b2d8aef90139869f927f4836a. + +The contiguous corner method, incidence factorization and exact partition +certificates are due to Rohan Arun (PR #31), method commit +02f68f95369c39955c5ffcfddbd75cd83e26b305. The selected (25,23) certificate +specializes Dominik Scholz's parameterization (PR #33), +a499a345c040aa15b68e984a0904eadc485c7a7e. Their source notices and proof +snapshots are preserved in references/copied-centers. Historical OpenAI +Codex, GPT-6 Astra and Anthropic Claude Opus 5.5 assistance is credited to +the respective imported sources. Existing PR #21/#23 and RaD attribution, +licenses and the Gaussian correction remain in force. + +Copied centers with reversed (23,25) data corners, 2026-10-08 +Rohan Arun with substantial OpenAI Codex assistance supplies the composition, +exact dimension23 controls and cost/assembly integration. James Chang's PR34 +supplies the reversed boundary family; icekylinx's PR36 supplies copied retained +centers and the selected complex graph. The balanced assembly remains attributed +to RaD/hipotures and its PR34 specialization. Retain Dominik Scholz PR33, Rohan +Arun PR31, Zhihao Chen PR29/21/23, Paureel/Aurel Prosz, Swapnil Jain, icekylinx, +eumemic, Douglas Colkitt, OpenAI and all earlier notices and hypotheses. +This is not a claim of formal verification, new foundational rank identities, +global optimality or measured runtime speedup. + +Fixed middle basis with copied reversed corners (2026-10-08): +Rohan Arun, with substantial OpenAI Codex assistance. Composition and +simultaneous-basis proof retain icekylinx PR32/36, James Chang PR34, +Dominik Scholz PR35/38, Zhihao Chen, Paureel, Swapnil Jain, RaD/hipotures, +eumemic and all preceding attribution and licenses. See research/copied-fixed-reversed. + +Changed-graph composition with copied centers and fixed bases, 8 October 2026 +Prepared with OpenAI Codex assistance. +The new composition uses icekylinx PR36 copied centers and PR32 fixed-I+J +projectors/profiler, Dominik Scholz PR35 two fixed factors and CRT argument, +Zhihao Chen (jacklightChen) PR29/23 interfaces, Aurel Prosz (Paureel) and +Swapnil Jain's two-stage development, Rohan Arun PR25/31 corner methods, +Dominik Scholz PR33/38 dimension/composition work; James Chang (jamesyc) +PR34 reversed geometry and balanced transfer; Rohan Arun PR37/39/40 +reversed integration, modular data tests and both-fixed baseline; +RaD (hipotures) PR41 changed scalar DAGs, alternating point order and +independent role compiler; RaD balanced routing/bulk interfaces, and all +retained predecessors. +Complete attribution and the conditional proof boundary are recorded in +research/copied-fixed/PROOF.md; imported originals and source pins are retained. +Original Apache-2.0 and CC0 notices remain applicable. + +Community follow-up integration (PRs #49, #45 and #26), 8 October 2026: +Rohan Arun contributes weighted carrier matching and climbed summand/leave-one-out +orders (PRs #44/#47/#49), with Anthropic Claude assistance. The final construction +composes Chafik Boukhalfa's changed-graph checks, exact rational data recovery and +reordered exclusion sums (PRs #43/#46/#48, OpenAI Codex assistance) with RaD / +hipotures's alternating pair order and independent physical role compiler (#41). +All predecessor attribution and conditional proof interfaces remain in force. + +Alejandro Zarzuelo Urdiales contributes Gaussian parity and canonical arithmetic +proofs, finite tensor execution, profile precision accounting and mixed-center +reference checks (#45), with substantial OpenAI Codex assistance. See +research/gaussian-parity-synthesis/PROVENANCE.md for his full attribution. +Ryan S / princezuda contributes historical Lean certificate arithmetic, algebraic +frame/movement/guard lemmas and independent paired-circuit checks (#26). Original +source notices and historical snapshots are preserved. Neither formal package +claims a full formal proof of the multiplication algorithm or a new kappa. + +Joint frame compilation of skip-prefix strips + +The local compiler and exact finite witnesses were prepared by eumemic with +substantial OpenAI Codex assistance. Avi Eisenberg / ikeboy with Anthropic +Claude assistance supplied the unmodified PR #53 skip-prefix producer. +The pinned PR #48 pipeline is Chafik Boukhalfa's contribution, incorporating +Rohan Arun, RaD / hipotures, icekylinx, Dominik Scholz, James Chang, +Zhihao Chen, Aurel Prosz / Paureel, Swapnil Jain, eumemic, Douglas Colkitt, +OpenAI, David Harvey, Joris van der Hoeven and the earlier contributors +credited in the retained sources. See references/frame-compiler/pr48/NOTICE +and references/frame-compiler/pr53/NOTICE for their original attribution. +No authorship or review by those predecessors is claimed for this compiler. + +Joint dual-skip/frame-compiler composition (2026-10-08) +--------------------------------------------------- +Rohan Gupta / gupt1156 with Anthropic Claude assistance: PR55 dual-skip +scalar graph; original LICENSE and NOTICE are retained under +references/frame-compiler/pr55. eumemic with substantial OpenAI Codex +assistance: PR57 joint reversible frame compiler, serialized physical-word +replay and profile reconstruction. Chafik Boukhalfa with OpenAI Codex +assistance: this composition, independent integration and exact certificate. +Avi Eisenberg / ikeboy, Rohan Arun, RaD / hipotures, icekylinx, and all +predecessor authors retain the notices above and in pinned source files. +No original copyright or Apache-2.0 notice is removed. + +Ranked reclamation after the PR58 joint-dual composition (2026-10-08) +---------------------------------------------------------------- +Chafik Boukhalfa with OpenAI Codex assistance: prioritize retired physical +slots by descending current frame rank, retaining slot ID as the tie-breaker; +complete replay, exact profiles, comparison and adversarial controls. +The dedicated joint_dual_reclaim_compiler.py is derived from eumemic's PR57 +compiler. Its original shared compiler remains byte-identical. The PR55 +producer and all preceding attribution remain unchanged. The complete PR58 +comparison certificate and its original source/proof provenance are retained +under references/frame-compiler/pr58. + +Skip-prefix strip producer, 8 October 2026 +By Avi Eisenberg (ikeboy), with Anthropic Claude (Claude Opus 5.5) +assistance. The skip-prefix leave-one-out strip layout in +research/skip-strips/skip_graph.py is new. Everything else reuses Chafik +Boukhalfa's PR #43/#46/#48 research/copied-fixed pipeline, its verification +code, exact data recovery and support-order sorting, together with: +- Rohan Arun's PR #44 weighted carrier matching optimizer; +- RaD (hipotures) PR #41 alternating point order; +- every predecessor credited above. +Complete attribution and the conditional proof boundary are recorded in +research/skip-strips/PROOF.md. Original Apache-2.0 and CC0 notices remain +applicable. + +Local split-pair skip-prefix research continuation (2026-10-08) +Prepared with substantial OpenAI Codex assistance at the user's request. +New contribution: singleton/pair recursive partition refinement, finite summand +order and moment search, and verified composition with inherited paid clones. +Avi Eisenberg/ikeboy with Anthropic Claude assistance: PR53 skip-prefix strips. +Chafik Boukhalfa with OpenAI Codex assistance: PR54 original-envelope clone +specialization and PR43/46/48 checking interfaces. RaD/hipotures: PR51 +whole-chain/source-partition cloning and earlier producer/interfaces. +Rohan Arun: PR44 weighted matching and earlier fixed/reversed work, with all +stated historical AI assistance. All predecessor notices remain applicable. + +Original-envelope whole-chain clones on skip-prefix strips (research/skip-clones): +Chafik Boukhalfa, prepared with OpenAI Codex assistance. The complete-chain +source-partition clone operation is adapted from RaD / hipotures, PR #51, +with original envelopes retained here. The scalar base is Avi Eisenberg's +PR #53 skip-prefix layout, prepared with Anthropic Claude assistance. +All credited PR #43/#46/#48, PR #44 and earlier code/proof notices remain +in force. This extension is provided under the repository's Apache-2.0 terms. + +Community review checkpoint #50–#61, 8 October 2026 +Maintainer integration and independent arithmetic: Douglas Colkitt, with OpenAI Codex. +Avi Eisenberg / ikeboy: skip-prefix strip framework (#53), with Anthropic Claude. +Rohan Gupta / gupt1156: parallel order search (#50, stated Antigravity assistance) +and dual-suffix strips (#55, Anthropic Claude assistance). +Eumemic: equal-frame joint compilation and paid signal reclamation (#57), with Codex. +Chafik Boukhalfa: original-envelope paid clones (#54), composition (#58) and ranked +reclamation (#60), with Codex. Rohan Arun: order search (#52) and positive-frame +composition (#56), with Anthropic Claude and stated Codex validation assistance. +RaD / hipotures: enlarged frames, source partitions and paid whole-chain clones +(#51), with Codex. Rohan Garg / rohangar1: singleton/pair recursion and paid-clone +composition (#59), with Codex. Alejandro Zarzuelo Urdiales: exact parameter +refinement, scoped Lean arithmetic, matrix/parity and search tools (#61), with +his disclosed multi-tool and Codex assistance. +Parallel #50/#52 records are reviewed and credited without replacing the earlier +climbed-48 implementation; their exact commit history is retained. The separate +#51 witness is retained as research; its source-graph/negative-basis data proof +has not received the same full replay as #56's consumed compiler machinery. +See docs/research/community-round2-review.md for precise validation boundaries. +All earlier licenses, contributions and notices remain applicable. + +Interval strips and core-aware pair assembly + +The interval leave-one-out strips and the core-aware group-pair assembly in +research/pair-assembly were prepared by Avi Eisenberg / ikeboy with Anthropic +Claude assistance. They build on PR #53's skip-prefix pipeline and its +inherited PR #48 sources. The stacked witness in research/pair-assembly/frame +uses eumemic's PR #57 joint frame compiler unchanged; its driver and verifier +are adaptations of PR #57's skip_frame_compiler.py, skip_frame_compose.py and +verify_skip_frame.py. Original notices are retained. + +Stopped product-ring extension, 9 October 2026 +Copyright 2026 icekylinx + +The opposite-bank generic projector factorization, componentwise atom-ring +adapters, stopped recurrence and outer ordinary-interchange conversion, +aligned pair-star producer with rational centers, exact denominator-21 +grid and selected (23,23)/(24,24) composition are contributed by icekylinx +with substantial OpenAI GPT-6 Astra and Codex assistance. This extension +builds on merged PR #36, 11817ccacb564bb7f98789c20dc11d3fece207e3, and is +integrated on upstream main 56b66d58297deca1d7dd130247d720e960f77a37. + +The retained ordinary leaf uses the preceding projector, corner and +copied-center proofs. The selected two-stage topology retains Aurel Prosz +(Paureel), Swapnil Jain and Zhihao Chen's credits; the semantic assembly +retains PR #21/#23 and RaD's attribution. Paired producers, carrier matching, +original source notices and each imported license remain unchanged. + +Partial-gauge extension, 9 October 2026 +Copyright 2026 icekylinx + +The pair-first cube assembly, compatible backward frame enlargement, +partial source-gauge schedule and stopped whole-projector composition are +contributed by icekylinx with substantial OpenAI GPT-6 Astra and Codex +assistance, extending PR #104 at 948ce1510df750f4c18b96bdaef436a86f8bf834. + +The physical deferred bit word, reflected event accounting, fan-order +clarification and signed two-stage endpoint are adopted from Zhihao Chen +(jacklightChen), PR #97, e15350b66e4bf796153a68c240a6cc1b788abe3e, with +Codex assistance. Its deferred readouts, V leaves, lifted frames and frozen +witness originate in Swapnil Jain's 741e7aa078392553815df7926ee17ac5e25a8c38, +with the original Claude disclosure retained. Cyclic interval and core-aware +pair primitives build on Avi Eisenberg (ikeboy), PR #62, +ad0f25ff7b23cff7f08ad237c2254e6ecf74257e, with Anthropic Claude assistance. +Prior two-stage, copied-center, ordinary-leaf, semantic/bulk, routing and +precision credits remain in force. Imported source notices and licenses +are preserved under references/partial-gauge/; see SOURCES.json. + +Three-stage cover extension, 9 October 2026 +Copyright 2026 icekylinx + +The arbitrary-subspace Clifford frame theorem, three-stage Cayley geometry, +weighted q-adic GL cover, unit-lower weighted compiler, charged local-ring +fallback, uniform batching, internal row borrowing and exact composition +are contributed by icekylinx with substantial OpenAI GPT-6 Astra assistance. This +repository integration uses Codex assistance and extends +PR #115, cf60442983c153a3a0f5272ea4713f99eba13c86. + +The selected local complex DAG and support/label replay are adopted from +Copyright 2026 eumemic, PR #117, cbb05ce504d571546d9b7794c186a613c659c3bf, +with its original Apache-2.0 terms and Claude assistance disclosure. +The unchanged bit physical word retains Zhihao Chen (jacklightChen), +PR #97, and Swapnil Jain's underlying witness and assistance notices. +The prior opposite-bank, stopped ordinary-leaf, copied-center, topology, +semantic/bulk, precision and analytic credits remain applicable. +Original source notices and license bytes are preserved in +references/three-stage-cover/pr117/ and references/partial-gauge/. + +Paired-cube and shared-core extension, 9 October 2026 +Copyright 2026 icekylinx + +The paired-coordinate triple-cube identity, signed overlap channels, +original-source involution, coordinate-star centers, concrete frame/carrier +compiler, chronological source gauges, PR97 gauge-subset selection, +orthogonal cross-stage routing and exact supplier/assembly integration +are contributed by icekylinx with substantial OpenAI GPT-6 Astra assistance. +Repository integration uses Codex assistance and extends PR #130 at +6a9970a530119174507904e23592fd59ede19a5d. + +The substantial completed-core workspace-sharing principle is credited to +an664, PR #128, 530588a019b4a74f09180680c9e3961bf649ec89, with its reported +OpenAI Codex assistance. No source code, partition or serialized witness +from that PR is imported. The new routing and paired-motif integration +are described separately and claim no exclusive priority for broad sharing, +paired-coordinate, Fourier or carrier-matching ideas. + +Selected coarse queries restrict the retained eumemic PR #117 positive DAG; +its original Apache-2.0 and Claude assistance notices remain unchanged. +The bit word and original frames retain Zhihao Chen's PR #97 and Swapnil +Jain's frozen witness, notices and historical assistance disclosures. +Earlier generalized Clifford, three-stage, uniform weighted recursion, +ordinary-leaf, semantic/bulk, precision and analytic credits remain in force. +Source-specific license bytes are preserved under references/. + +October 9 recycled-bit integration: +DaysSky (#150), William Porter/hpst3r (#147), Thomas Marchand/Th0rgal (#146), +Rohan Gupta/gupt1156 (#148), James Chang/jamesyc (#124), SovereignSteak (#122). +The original notices and assistance disclosures are retained under research/, +including research/recycled-bit-integration/NOTICE. Andrew Barnes/Bortlesboat +(#101) repairs shared verifiers using rfu08's #64 counterexample. rfu08's +scoped finite and formal-transfer packages are included separately; they do +not establish the full multiplication theorem. Earlier and parallel credits +remain applicable. Douglas Colkitt integrated these with OpenAI Codex assistance. + +SovereignSteak PR151 independently supplies the matching gauge/recycling +composition and the selected 473/10^6 atom refinement; its original package +and assistance disclosure are retained under research/gauge-recycling/. + +October 9 coordinated-frame and completed-entrance-bank integration: +Dugongue PR186, 166a34d75e853e933855f0f7e940fe77c4bb6b8d, builds on +Chafik Boukhalfa/chafreaky's PR181, 7fb2194801e3a10f54772c7f0d5505035a4fc0ad, +and eumemic's PR168 v4, 4a3c769e5c5430e7114c4d3e099ff34664677f17. +The frozen package and prerequisite archive preserve all author, source-license +and assistance notices, including PR175's compiler optimization. DaysSky's +carrier closure, James Chang's terminal substitution and balanced assembly, +geckods' stopping refinements and all preceding framework credits remain. +Rohan Arun's PR185 bootstrap and chafreaky's PR182 bit refinement are reviewed +parallel contributions; their gains are not added to the selected witness. +Douglas Colkitt integrated the package and supplied the bank scheduling +supplement with OpenAI Codex assistance, under Apache-2.0. The all-size +interfaces remain conditional; finite verification is not full formalization. diff --git a/research/relay1256-sharp-scalar-range/credits/OUTER-GATE-NOTICE b/research/relay1256-sharp-scalar-range/credits/OUTER-GATE-NOTICE new file mode 100644 index 0000000000..1bc9b4da2b --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/OUTER-GATE-NOTICE @@ -0,0 +1,13 @@ +# Attribution and source reuse + +New independent operator/invoice verification and fixed-grid interval implementation: OpenAI Codex-assisted research session, 2026-11-10. Apache-2.0. + +DreamingOfClouds supplied the #6936 native predecessor and #6937 shared-wire backup/E-free root. CubeJerry supplied #6912 caller blocks, #6980's QUERY-only caller and #6970/#6985 precision methods. Preserve Ali Romman's #5599 helper-split attribution, icekylinx's #104 odd-grid attribution and unpaidmoon's #6930 width-phase attribution in the parent construction. + +Three unchanged source files are authenticated by exact SHA256 and source commit/path in `INDEPENDENT-SOURCES.json`: + +- #6970 `independent_bounds.py`: fixed integer-grid and positive caller-transfer engine. +- #6985 `envelopes.py`: positive forward envelopes and backward covector witnesses. +- #6985 `source_binding.py`: retained helper functions for the envelope engine. + +The adapter applies only two literal helper-census substitutions in memory and records them. It does not call that adaptation an independent mathematical implementation. The new reverse native operator, formal caller ring and elementary interval engine do not import the source authors' operator or invoice checkers. diff --git a/research/relay1256-sharp-scalar-range/credits/PR7663-NOTICE b/research/relay1256-sharp-scalar-range/credits/PR7663-NOTICE new file mode 100644 index 0000000000..6ce1bc49a3 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/PR7663-NOTICE @@ -0,0 +1,7 @@ +Apache-2.0. Ali Romman with substantial OpenAI Codex assistance. + +The new material is the joint chart assignment, its source-bound operator and caller checks, and the fresh-cost complete-root scalar and private-bank specialization. The parent data are from Clear1nsight's PR #7660 on DreamingOfClouds' PRs #7658 and #7641 and SYLforge's PR #7637. The contributors credited in those sources retain their credit. Original notices are in source/parent-NOTICE, source/physical-NOTICE and source/word-NOTICE. + +Rohan Arun supplied the 17-stage signed caller in PR #7113. The unmodified interval gate and the exact caller data come from its pinned source. See source/caller-NOTICE and source/independent-NOTICE, and the source file inventory for all consumed bytes. The two operator engines and integer power evaluator come from the declared current-main inputs. + +Jerry Chen (CubeJerry) supplied selected PRs #7000 and #7001. The selected application and its phase, common-root, finite-cost and compiler arguments, the odd-native common-grid package, the prior typed six-Gamma reset of PR #7586 and the contributors credited in those records provided the foundations used here. PR #7661 independently derives the fixed caller residuals. PR #7662 supplies a rounded parent-clock alternative. This package uses neither implementation. diff --git a/research/relay1256-sharp-scalar-range/credits/PR7669-PROVENANCE.md b/research/relay1256-sharp-scalar-range/credits/PR7669-PROVENANCE.md new file mode 100644 index 0000000000..bbee2437f0 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/PR7669-PROVENANCE.md @@ -0,0 +1,43 @@ +# Provenance, credit and scope + +This is a cumulative contribution. Thank you to everyone who has made their constructions, exact inputs, failure cases and careful reviews available. Those contributions made this composition possible; the small increment here should not obscure the much larger body of work it rests on. We aim to leave equally inspectable material for the next researcher. + +## Direct lineage and reused methods + +- **SYLforge, #7637:** the H5B frame-restored233-CLEAR construction is the published ancestor of this research. The present K372-2 source retains its descendants rather than starting a new independent construction. [PR7637](https://github.com/CrocSwap/integer-mult-bounds/pull/7637). +- **DreamingOfClouds, #7641 and #7658:** #7641 combines20 edits over #7637. #7658 supplies the large-neighbourhood construction and its exact joint hub reconstruction: helper360/537 SWAP-NULL relabel, main176 bypass, helper236 contact re-chart, helper370 last-stay removal and helper461 first-stay removal. The source explicitly credits its #7630/#7636 relabel lineage and #7634 swap law. Our source descends from this word; those gains are not claimed as new work here. [PR7641](https://github.com/CrocSwap/integer-mult-bounds/pull/7641), [PR7658](https://github.com/CrocSwap/integer-mult-bounds/pull/7658). +- **Clear1nsight:** the contact re-chart type from #7612 and the24-bit coefficient-splitting report are part of the construction's lineage. #7626 contributed source isolation/portable materialization, and #7631/#7633 are credited by the parent lineage. #7660 separately published the role379 shortcut already present in our source; we do not count its rediscovery as another improvement. [PR7612](https://github.com/CrocSwap/integer-mult-bounds/pull/7612), [coefficient report](https://github.com/CrocSwap/integer-mult-bounds/discussions/7111#discussioncomment-19691406), [PR7660](https://github.com/CrocSwap/integer-mult-bounds/pull/7660). +- **ToikY, #7643:** the contact re-chart work is explicitly credited by #7658 alongside Clear1nsight's #7612. Its74-erasure contract and κ are not added to this candidate. No shared CI repair is copied by this PR. [PR7643](https://github.com/CrocSwap/integer-mult-bounds/pull/7643). +- **Rohan Arun, #7113:** the signed caller/accounting lineage underlies the actual16-call composition. Its original source commit is07a76c5e2738882c0cb322278e56f6cae3c9885d. We recount the literal16-call input here; we do not substitute historical17-call constants. [PR7113](https://github.com/CrocSwap/integer-mult-bounds/pull/7113). +- **Ali Romman and upstream contributors:** #7605 supplies an earlier exact protected/application checkpoint cited by the #7637 lineage. The two unmodified directed invoice comparison files are copied from canonical upstream base e54523517d0f03e58352e647c403a5af3e525a57, retaining their attribution and Apache-2.0 headers. Their arithmetic is not a full scientific acceptance gate. [PR7605](https://github.com/CrocSwap/integer-mult-bounds/pull/7605). +- **CubeJerry, Douglas Colkitt and the earlier source contributors:** the retained framework, primitive/service work and repository infrastructure form part of the inherited basis. This package does not claim that their open supplier obligations are resolved by a successful finite replay. + +## What this contribution adds + +The predecessor K372-2 supplies five EARLY edits (1188,1013,940,1141,1195, with940/1141 sharing one ADD). This construction recomputes the1517/770 bypass on that actual source and combined it with a three-field triangle repair. The new contribution is their compatible composition, exact12-block edit, final235 CLEAR witnesses and a small reproducible publication package. Substantial OpenAI Codex assistance was used for research and packaging. New work is Apache-2.0. + +Exact source canonical:3dd1199d95291ee96e6c0aeb6f55f86244a569b1d4975840bf0207e2dcf9c025. Exact candidate canonical:ee251f6005d2a92b887c5b804bd7aac05989d278b2c9ccb3c3091b50d533b08d. Byte hashes are in MANIFEST.json. The12 edit blocks specify every changed event and leave all other fields unchanged. The parent research generators are not silently imported; this publication uses the transparent exact edit as its deterministic reconstruction recipe. + +## Related work, not automatically incorporated + +Ali Romman's #7662 studies one half-ADD rounding-error propagation with CubeJerry's service contract. Nick Petalas's #7654 provides a scoped minimal-rank interpolation result used as background when assessing structural repairs. Neither is a substitute for this candidate's entire rounded caller or physical supply proof. Their actual domains and assumptions must be checked before application. [PR7662](https://github.com/CrocSwap/integer-mult-bounds/pull/7662), [PR7654](https://github.com/CrocSwap/integer-mult-bounds/pull/7654). + +At the publication checkpoint, upstream main remained e54523517d0f03e58352e647c403a5af3e525a57. Ali Romman's #7663 uses a separate17-call full-private-bank model and changed helper endpoints; Clear1nsight's #7664 concerns180erasures/17calls; #7666 charges bank selection and return without claiming a new κ. K372-4 is later, separate research with changed helper-boundary conditions. None is silently folded into ee251f60 or used to assert identical physical assumptions. This is not a claim to the greatest bound across all models. + +## Fresh versus reused evidence + +Publication replay freshly reconstructs the whole word, runs exact rows/CLEAR, recounts native/caller histograms and checks the four directed price points. Normal, optimized Python and a Git-free committed-byte export were tested. The recorded ordinary/precision/native-phase/owner evidence retains its narrower scope and original source identifiers; it is not advertised as freshly rerun. No selected record, frozen predecessor package or common verifier is changed. + +## Editorial normalization of publication labels + +The follow-up editorial commit replaces workstation labels with construction terminology. It changes no candidate, edit, certificate, invoice, numeric result, executable code or contributor attribution. The original recorded bytes remain retrievable at commit `95d476623f1468822057554475dc8badb19bfa44`. Evidence JSON field labels and descriptive scopes are normalized, not recomputed scientific results. In NATIVE-EVIDENCE-REUSE.json, `predecessor-receipt:K372-2/NATIVE-NEW.json` is a historical receipt identifier, not a packaged file path. In PRICE-SUMMARY.json, recorded command invoice arguments now use the equivalent packaged `data/INVOICE.json` path; this is a portability transcription, not the verbatim original command. The fresh replay constructs and runs its own package-relative command. MANIFEST.json authenticates the edited publication bytes. + +Evidence byte-custody mapping (original -> editorially normalized SHA-256): + +- `evidence/BUILD.json`: `4744201691120223ded62a3e0d55ea1029203f3ea996305163e7db8d9e3679a4` -> `a0bafbb3f44920ed6ac1ed3d83385f025f5bbbcf5e9c8013ab1ca55ccf23c32d`. + +- `evidence/NATIVE-EVIDENCE-REUSE.json`: `6615b8a051ac9e4d8c73b5decf8a70d2a33d14da0c4c369574dec946984aa351` -> `84bab14d5c584aa8853b8e1025b562541ac6657e458d120ec803f2bc149c3eb8`. + +- `evidence/ROLE770-BUILD.json`: `c4ee72317678791d4b801479df0af052f3a50662c11ce3f3393b2151368fc92d` -> `1c400f16efbc427c40c811a57dc656fe97c2c84f94c71529e2c5c0598b146369`. + +- `evidence/PRICE-SUMMARY.json`: `c2777382c677634f7017b68c55dbd831a40945b77057f26b738f857c17ee8759` -> `d13741fe3217a01edd0b79b40dad037e644e441fbbc300838df99c0b050a5f42`. diff --git a/research/relay1256-sharp-scalar-range/credits/PR7674-NOTICE b/research/relay1256-sharp-scalar-range/credits/PR7674-NOTICE new file mode 100644 index 0000000000..951f081717 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/PR7674-NOTICE @@ -0,0 +1,7 @@ +Apache License 2.0. Research prepared for Ali Romman with substantial OpenAI Codex assistance. + +New work includes the fifteen certified sender/receiver pivots, the forward and transposed communication construction, the complete native row checker, and the finite stock and precision application for this 16-call program. + +SYLforge supplied the composition, native word reconstruction, published invoice and exact 16-stage caller bytes from PR #7669. Their original PROVENANCE.md is preserved in source/7669-CREDITS.md. It credits the H5B work of SYLforge, DreamingOfClouds' joint native reconstructions, Clear1nsight's coefficient splitting, role-379 and helper work, Rohan Arun's signed caller lineage, and prior contributions from CubeJerry, Douglas Colkitt and others. The source/ directory holds the exact file bytes identified in SOURCE-PINS.json. + +The group-ring caller checker, independent operator checks, bit service, finite-cost method and field guide adapt the corresponding code and arguments of Ali Romman's PR #7663. That package in turn builds on the contributors named in its preserved source/7663-NOTICE. The unchanged directed log/exp gate retains its listed sources in source/gate-NOTICE. Earlier paid-transfer and bank-support constructions include PRs #7434 and #7576. Jerry Chen (CubeJerry) supplied the selected PRs #7000 and #7001 and the typed initializer in PR #7586. The per-call odd-grid induction and the selected application review provide the complete-root realization and finite-cost substitution. Douglas Colkitt and the credited repository authors supplied the pinned operator engines, price evaluator and research workflow. The license accompanies this package. diff --git a/research/relay1256-sharp-scalar-range/credits/PR7682-NOTICE b/research/relay1256-sharp-scalar-range/credits/PR7682-NOTICE new file mode 100644 index 0000000000..4654631de0 --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/PR7682-NOTICE @@ -0,0 +1,9 @@ +Apache License 2.0. Research prepared for Clear1nsight with substantial OpenAI Codex assistance. + +Ali Romman supplied the fifteen exact banked pivots and the frozen #7674 baseline, along with the complete signed 16-call graph and application method. The original notices remain at peer/NOTICE, peer/source/7663-NOTICE and peer/source/gate-NOTICE. Read peer/SOURCE-PINS.json for each ancestry binding and ASSETS.json for the exact vendored bytes at scientific commit 26b6c469aa6c4aa04322ce25a22150f90e5fb14e and refreshed head 28b2af322a1fab01b3734365988930920f9dda48. + +SYLforge supplied the #7669 composition, native source pins and literal caller. Its preserved parent provenance is peer/source/7669-CREDITS.md. Retain DreamingOfClouds' #7641/#7658 lineage, Clear1nsight's integer scalar/contact work including #7612, ToikY's #7643 contact-rechart precedent, Rohan Arun's #7113 signed-caller lineage, 5080's helper work, and all other contributors named in the original records. Ali Romman's #7663 exact-grid and extent proof and #7666/#7670 selection and guide work supply relevant implementation precedents. Jerry Chen (CubeJerry) supplied the selected #7000/#7001 applications and #7586 typed initialization. Douglas Colkitt and the authors in the original notices supplied the frozen native compiler, independent checker, price evaluator and package infrastructure. + +The release base reuses ToikY's exact shared CI and historical replay sync 0eabc624a9bc9ed2847719efd63390db4d297df4 above canonical e54523517d0f03e58352e647c403a5af3e525a57. This workflow provenance is separate from the banked construction and its mathematical comparison. + +This package credits the original role-230 splice and its same-contract transfer to the unchanged fifteen-pivot graph, the independent exact wire and price replays, matched complete resource accounting and source-bound native and caller verification. The original works retain their authorships and licenses. diff --git a/research/relay1256-sharp-scalar-range/credits/PR7704-NOTICE b/research/relay1256-sharp-scalar-range/credits/PR7704-NOTICE new file mode 100644 index 0000000000..7d2dfec60a --- /dev/null +++ b/research/relay1256-sharp-scalar-range/credits/PR7704-NOTICE @@ -0,0 +1,5 @@ +Apache License 2.0. Prepared for Clear1nsight with substantial OpenAI Codex and ChatGPT Pro assistance. + +SYLforge supplied the cf6a nine-swap word and its normalized verification in #7700. DreamingOfClouds contributed the #7641 and #7658 swap searches. Ali Romman supplied the changed helper contract in #7663 and the charged bank construction in #7674. Rohan Arun and the #7113 lineage supplied the signed caller. The bank, native, physical and directed price methods reuse work by Ali Romman, SYLforge, Clear1nsight and the contributors credited in #7669, #7682 and #7697. Jerry Chen (CubeJerry) supplied the selected #7000 and #7001 application. Douglas Colkitt and the recorded authors supplied the selected framework, compiler, checker and package infrastructure. + +This contribution identifies and contracts the two paid transfers on helper 1256, reconstructs the banked graph from the supplied source, and checks the resulting physical application and complete signed price. The separate investigation and technical review were performed with Codex agents. Preserve the original notices and citations in credits/ when redistributing the source data or derived engines. See SOURCES.json for pinned upstream inputs and ASSETS.json for file-level custody. diff --git a/research/relay1256-sharp-scalar-range/data/CANDIDATE-GRAPH.json.gz b/research/relay1256-sharp-scalar-range/data/CANDIDATE-GRAPH.json.gz new file mode 100644 index 0000000000..814a8f004d Binary files /dev/null and b/research/relay1256-sharp-scalar-range/data/CANDIDATE-GRAPH.json.gz differ diff --git a/research/relay1256-sharp-scalar-range/verify.py b/research/relay1256-sharp-scalar-range/verify.py new file mode 100644 index 0000000000..a5ac9360bf --- /dev/null +++ b/research/relay1256-sharp-scalar-range/verify.py @@ -0,0 +1,198 @@ +"""Check sharp scalar ranges on the fixed PR7704 normalized graph.""" +from fractions import Fraction +from pathlib import Path +import gzip +import hashlib +import json + + +ROOT = Path(__file__).resolve().parent +GRID = 921600 +GRAPH_HASH = '8e7b755bd2d2d387bc3d3adfe48100985adf74775ca5acbb143170a3d16b5ccf' +ARCHIVE_HASH = '0177c25371181e3b1f2c13325eb8434fbf6384a9ab99a612849efd5d0f6ab2cc' +KINDS = ('snapshot', 'prefix', 'product', 'numerator') + + +def require(condition, message): + if not condition: + raise ValueError(message) + + +def encode(value): + return json.dumps(value, sort_keys=True, separators=(',', ':')).encode() + + +def read_graph(): + archive = (ROOT / 'data/CANDIDATE-GRAPH.json.gz').read_bytes() + require(hashlib.sha256(archive).hexdigest() == ARCHIVE_HASH, 'source archive changed') + raw = gzip.decompress(archive) + require(hashlib.sha256(raw).hexdigest() == GRAPH_HASH, 'source graph changed') + graph = json.loads(raw) + forms = {key: [(donor, Fraction(q)) for donor, q in row] + for key, row in graph['forms']} + moves = {key: (start, end) for key, start, end in graph['moves']} + n = len(graph['initial']) + terminals = set(range(graph['first_output'], graph['first_output'] + n)) + require(n == 1501 and len(moves) == 6143 and len(forms) == 7644, + 'source dimensions changed') + require(set(forms) == set(moves) | terminals and not terminals.intersection(moves), + 'row namespace changed') + require(len(graph['schedule']) == len(forms) and set(graph['schedule']) == set(forms), + 'schedule is not a permutation') + charts = {~column: chart for column, chart in enumerate(graph['initial'])} + for key in graph['schedule']: + start = moves[key][0] if key in moves else graph['final'][key - graph['first_output']] + require([d for d, _ in forms[key]] == sorted({d for d, _ in forms[key]}), + 'literal donor order changed') + require(all(q and d in charts and charts[d] == start for d, q in forms[key]), + 'donor is absent, zero or in a different chart') + if key in moves: + charts[key] = moves[key][1] + return graph, forms, moves + + +def basis_ranges(graph, forms, moves): + values = {~column: {column: GRID} for column in range(len(graph['initial']))} + peaks = {kind: {'mass': GRID, 'row': None, 'term': None, 'vector': {0: GRID}} + for kind in KINDS} + triangle = {'mass': GRID, 'row': None} + trace = hashlib.sha256() + products = prefixes = 0 + terminal_peak_row = None + + def record(kind, mass, key, term, vector): + if mass > peaks[kind]['mass']: + peaks[kind] = {'mass': mass, 'row': key, 'term': term, + 'vector': dict(vector)} + + for key in graph['schedule']: + row = {} + mass = row_triangle = 0 + for term, (donor, q) in enumerate(forms[key], 1): + product = {} + for column, value in values[donor].items(): + z, remainder = divmod(q.numerator * value, q.denominator) + require(remainder == 0, 'literal product is off the declared grid') + product[column] = z + old = row.get(column, 0) + new = old + z + mass += abs(new) - abs(old) + if new: + row[column] = new + else: + row.pop(column, None) + products += 1 + product_mass = sum(map(abs, product.values())) + row_triangle += product_mass + record('product', product_mass, key, term, product) + record('numerator', product_mass * q.denominator, key, term, + {column: z * q.denominator for column, z in product.items()}) + record('prefix', mass, key, term, row) + prefixes += 1 + record('snapshot', mass, key, len(forms[key]), row) + if row_triangle > triangle['mass']: + triangle = {'mass': row_triangle, 'row': key} + trace.update(encode([key, sorted(row.items())]) + b'\n') + if key in moves: + values[key] = row + if key == 156140: + terminal_peak_row = sorted(row.items()) + require(products == 9044092 and prefixes == 172414, 'source coverage changed') + require(terminal_peak_row == [(1372, GRID)], 'peak row no longer cancels to one input') + require(Fraction(triangle['mass'], GRID) == Fraction(13819451, 16), + 'source triangle bound changed') + return peaks, {'rows': len(forms), 'ordered_prefixes': prefixes, + 'nonzero_basis_products': products, + 'completed_rows_sha256': trace.hexdigest(), + 'separate_product_triangle_maximum': str(Fraction(triangle['mass'], GRID)), + 'triangle_peak_row': triangle['row'], + 'peak_prefix_row_final_value': 'input 1372'} + + +def scalar_oracle(graph, forms, moves, inputs, target): + """Use scalar rational arithmetic, with no basis vectors or range recurrence.""" + values = {~column: Fraction(value) for column, value in enumerate(inputs)} + maxima = {kind: max(map(abs, inputs)) for kind in KINDS} + observed = None + integral = True + for key in graph['schedule']: + total = Fraction() + for term, (donor, q) in enumerate(forms[key], 1): + numerator = values[donor] * q.numerator + product = numerator / q.denominator + total += product + for kind, value in [('numerator', numerator), ('product', product), ('prefix', total)]: + maxima[kind] = max(maxima[kind], abs(value)) + integral = integral and value.denominator == 1 + if (kind, key, term) == target: + observed = value + maxima['snapshot'] = max(maxima['snapshot'], abs(total)) + if ('snapshot', key, len(forms[key])) == target: + observed = total + if key in moves: + values[key] = total + require(observed is not None, 'oracle target was not visited') + return maxima, observed, integral + + +def extra_bits(mass): + bits = 0 + while mass >= GRID << bits: + bits += 1 + return bits + + +def check(): + graph, forms, moves = read_graph() + peaks, coverage = basis_ranges(graph, forms, moves) + bounds = {kind: Fraction(peak['mass'], GRID) for kind, peak in peaks.items()} + witnesses = {} + width = 32 + magnitude = ((1 << (width - 1)) - 1) // GRID * GRID + for kind, peak in peaks.items(): + signs = [-1 if peak['vector'].get(column, 0) < 0 else 1 + for column in range(len(graph['initial']))] + target = (kind, peak['row'], peak['term']) + maxima, observed, _ = scalar_oracle(graph, forms, moves, signs, target) + require(observed == bounds[kind], 'corner does not attain its bound') + require(all(maxima[name] <= bounds[name] for name in KINDS), 'oracle exceeds a bound') + scaled, integer_value, integral = scalar_oracle( + graph, forms, moves, [magnitude * x for x in signs], target) + require(integral, 'integer-lattice witness is not exact') + require(integer_value == magnitude * bounds[kind], 'scaled corner changed') + growth = extra_bits(peak['mass']) + require(all(value < 1 << (width + 21 - 1) for value in scaled.values()), + 'W+21 data range failed') + require(integer_value >= 1 << (width + growth - 2), 'one-bit-smaller control did not fail') + donor, q = forms[peak['row']][peak['term'] - 1] + witnesses[kind] = { + 'maximum_for_unit_inputs': str(bounds[kind]), 'row': peak['row'], + 'one_based_term': peak['term'], 'source_term': [donor, str(q)], + 'support': len(peak['vector']), 'coefficient_vector_sha256': + hashlib.sha256(encode(sorted(peak['vector'].items()))).hexdigest(), + 'negative_input_columns': [i for i, sign in enumerate(signs) if sign < 0], + 'all_other_input_signs': 1, 'sufficient_uniform_extra_signed_bits': growth, + 'input_signed_width': width, 'input_integer_magnitude': magnitude, + 'attaining_integer_value': int(integer_value), + 'one_bit_smaller_positive_limit': (1 << (width + growth - 2)) - 1, + } + require([witnesses[k]['sufficient_uniform_extra_signed_bits'] for k in KINDS] + == [15, 18, 18, 21], 'bit bounds changed') + maximum_denominator = max(q.denominator for row in forms.values() for _, q in row) + require(maximum_denominator == 384, 'division remainder allowance changed') + return {'status': 'PASS_RELAY1256_SHARP_SCALAR_RANGE', + 'source_graph_sha256': GRAPH_HASH, 'integer_entrance_multiple': GRID, + 'coverage': coverage, 'sharp_ranges': witnesses, + 'scalar_oracle_complete_runs': 8, 'one_bit_smaller_controls': 4, + 'maximum_coefficient_denominator': maximum_denominator, + 'long_division_remainder_strict_upper': 2 * maximum_denominator, + 'source_triangle_to_sharp_prefix_ratio': str(Fraction(13819451, 16) / bounds['prefix']), + 'scope': 'Fixed normalized scalar graph; no physical chart or complete kappa claim.'} + + +if __name__ == '__main__': + result = check() + require(result == json.loads((ROOT / 'RESULT.json').read_text()), 'receipt differs') + print(result['status']) + print('7644 rows; 172414 prefixes; 9044092 basis products; 8 scalar oracle runs') + print('Extra signed bits: snapshots 15; prefixes 18; products 18; numerators 21')