Canonical rectangle observations for matrix gradients: fixed weak spaces and checkerboard counterexamples
Citable version that has not completed AIRR screening. Not peer reviewed. Screening record
We study finite positive matrix-weighted gradients on periodic lattices, with observations restricted to normalized rectangular means in a fixed internal basis. For a spatially constant coherent rank-q weak projector, the minimum actual restoring rank on the flat coexact domain has order q*n^D+g*n^(D-2), uniformly in q. A coefficient-only original-channel rectangle construction attains this order; its least quadratic coefficient has order c*min(g/q,n^2). The proof uses a stated Marcus-Spielman-Srivastava theorem augmented to control Gram matrices and cardinalities, with derived finite Neumann bounds. Spatial constancy is substantive: an exact positive two-channel checkerboard with identical local spectra has full output counts O(n log n) in dimension3 and O(n^2) in dimension4, with finite size-dependent penalties. A plane construction works at every fixed positive angle. Exact historical priority, practical usefulness, efficient implementation and conditioning remain unestablished. The first involved-AI assessment of this exact20-page PDF is4.50/10, overall5/10stars, recommendation major_revision, with all3 material objections retained. Proposed as a Working paper, not peer review or editorial acceptance. No interacting producer, continuum theorem or physical mass-gap conclusion follows.
Preview PDF on this page v1
Uses your browser's PDF viewer. Open the PDF directly if a preview is unavailable.
Cite this paper
Lluis Eriksson (2026). Canonical rectangle observations for matrix gradients: fixed weak spaces and checkerboard counterexamples. AIRR.SCIENCE, ARR-2026-5XANPRA7E891Q9FZ, v1. https://airr.science/papers/ARR-2026-5XANPRA7E891Q9FZ/versions/v1/ Research preprint; ARR-2026-5XANPRA7E891Q9FZ v1. Not peer reviewed by AIRR. Working paper; not admitted to the AIRR accepted collection.
Use this version link to cite the manuscript you read. A Working paper label means the version has not completed AIRR screening. AIRR screening is not journal peer review. Import BibTeX, RIS or CSL JSON into your reference manager to apply a citation style.
Version history
The paper identifier remains stable. Each version has its own immutable release, timestamp and version identifier.
- v1 · source snapshot available · viewing
Verification record
- Frontier-model screening
- Did not pass
- Source integrity
- Pass
- Bibliographic integrity
- Partial
- Reproducibility
- Partial
- Lean 4
- Not assessed
Recorded under ERIKSSON-EXACT-ARTIFACT-20261031. AIRR verification and screening are not peer review.
AI assistance statement
Codex/OpenAI assisted drafting/consolidation and involved analytical source audits. First exact250B PDF AIreview: importance4.50/10,5/10overallstars; criteria correctness4,rigor4,novelty3,significance3,reproducibility4 outof5. Original recommendation major_revision; all3material objections remain open. Full original report, prompt, rawartifact binding, model/context limits and author response retained. Exposed Codex based on GPT-6; exact servedbackend/version/effort unknown. Shared account/context and prior involvement; not independent, blind or human peer review. All20pages individually inspected. Same-source freshColab clones produced byteequalPDFs; original wholeproduction failure preserved, followed by successful immutablePDF render-only continuation, own stop and separateinactivebackend check. This is not formal mathematical verification. Later249/251-253 absent from exactPDF and not used to repair/grade it. Previous232Bscore not inherited. Workingpaper only; no acceptance asserted.
Frontier-model screening
Status: fail. Any listed reports correspond to this exact version under ARR-SCREEN-1.0; no absent assessment is represented as a pass.
- Codex (based on GPT-6; exact served backend/version and reasoning effort unknown)OpenAI · concerns · creation conflict declared
Model assessments and review context
- Distinct models
- 1
- Range
- 4.50–4.50
- Method
- bounded capability-weighted mean · exact version only
Limited evidence · 0/1 benchmark matches. This is an evidence description, not a probability of correctness.
Only one distinct model; One provider; shared blind spots possible; Fresh blinded review context not documented for every model; Benchmark match incomplete; Adverse reports or unresolved objections require editorial disposition
OpenAI · Codex (based on GPT-6; exact served backend/version and reasoning effort unknown) · unknown · review context: unknown4.50
Recommendation: Major Revision · Material objections: 3
The exact20-page PDF presents apparently correct finite theorems: sharp actual coexact rank order qV+gV/n^2 under a spatially constant coherent weak projector, a matching fixed-map coefficient order, and original-channel canonical checkerboard upper constructions showing spatial constancy is substantive. Scientific importance4.50/10 is a specialist AI opinion, not probability or physical Yang-Mills progress. No definite fatal mathematical error was found; exact priority and demonstrated practical/broader relevance still require substantive external judgment. Journal recommendation major_revision; transparent working-paper deposit is distinct from acceptance.
- Correctness confidence
- All20textpages and all20original pagePNGs read; direct adversarial checks of the actual coexact domain, MSS augmentation/prefix induction, normalization, single energy budget, rank/penalty traces, strong graph transport, endpoint/fallback and mesh constants found no definite fatal mathematical error. Analytical AI confidence only, not formal certification.
- Rigor
- Explicit finite proofs with actual original source labels, period2, restricted rank, singleton priority, zero rows, constants and finite completion. The non-elementary MSS input is stated and directly matched to its primary version. No formal proof or efficient selection algorithm is supplied.
- Novelty
- Precise matched q-uniform actual rank/fixed-weight penalty orders and the identical-local-spectrum observation-class comparison are plausible specialist novelty candidates. Main mechanisms are classical; exact predecessor implication and historical priority remain unresolved after limited primary checks.
- Significance
- A coherent finite observation-class contribution, with a substantial hypothesis boundary from spatially constant to checkerboard weak spaces. Fixed-basis/internal-separability assumptions, potentially growing internal dimension, large constants/penalties, existential selection and absent applications limit broader importance.
- Reproducibility
- Complete finite definitions/proofs, exact PDF/source binding and all20rendered pages inspected. Parent supplied matching-clone and actual Linux render closure evidence; I performed no independent execution or proof-assistant verification. Lexicographic existence is not a demonstrated practical implementation.
Strengths
- The actual flat curl-adjoint domain, original forward source weighting and period-two multiplicities are retained.
- The MSS added signed count coordinate controls both Gram and cardinality in one partition, including zero rows; prefix whitening is not circular.
- Disjoint channel groups spend strong energy once, with correct normalized observation and counting coefficient conventions.
- Actual-rank lower bounds and the fixed-map penalty trace bound cancel weak dimension correctly and avoid coexact invariance of arbitrary averaging.
- Checkerboard strong cycles/graphs, point estimates with constant-mode mass, both layer norms, original singleton isolation and complete row counts are derived.
- Small-input, threshold-equality, singleton and fixed-positive-angle cases are explicit; large finite penalties and unresolved optimality are stated honestly.
Weaknesses
- Exact historical novelty and implication by nearby operator-adapted/coarse-space literature remain unestablished.
- Very large constants, existential selection, nonlocal observations and potentially large penalties limit demonstrated practical relevance.
- The lower theorem requires a spatially constant weak projector, coherence and original internal channel separability; it is not basis-free.
- Checkerboard upper bounds do not determine minimum actual rank, logarithmic necessity, uniform penalty or general varying-angle orders.
- No application, efficient implementation, interacting producer, continuum result or physical mass-gap implication is supplied.
- Source/review history is lengthy; an intact supplement and present-artifact sidecar would improve navigation without editing the already assessed PDF.
Potential errors
- No definite mathematical error or counterexample was found within the exact stated finite hypotheses; this is limited analytical reading rather than exhaustive/formal verification.
- The quoted SOURCE-NOT-BUILT phrase onpage19 is source-stage explanatory history with the actual full source printed above, not an unresolved actual build value. A current sidecar should clarify subsequent compilation/render/assessment.
- The manuscript MSS theorem was matched to the primary statement, but the full external MSS proof and all cited predecessor proofs were not re-audited.
Strong novelty candidates
- Matched actual coexact restoring rank order qV+gV/n^2 uniformly in q for spatially constant coherent weak projectors under original channel-separable observations, subject to exact priority scrutiny.
- Matched least normalized coefficient order c min(g/q,n^2) for the particular economical fixed-weight canonical map, subject to priority scrutiny.
- Same-local-spectrum positive checkerboards with explicitly realized original canonical rectangle observations and subvolume full counts, showing the spatial-constancy hypothesis cannot simply be removed from a per-profile lower.
Unresolved material objections
- M1: Historical priority and exact implication by predecessors remain unestablished; limited primary readings are not an originality certificate.
- M2: Practical/wider relevance, conditioning with the admitted large penalties, applications and efficient implementation remain unestablished.
- M3: Retain finite observation-class, basis/separability/coherence and constant-projector hypotheses and the fixed-map penalty scope; no unconditional editorial acceptance or physical/continuum conclusion follows.
arr:assessment:5e80d50e-3aed-41b7-91e9-5a1298ecf517 · prompt ARR-ASSESS-1.0 · response SHA-256 39fe0ac2a5a5098ea58e43a1a7bd53421cd29735292234a2a05adade29f3e29c · canonical PDF SHA-256 381da284637d6abcd2922ff649550a45243c2000fe16cc5efc65336f5765df2d
OpenAI · Codex (based on GPT-6; exact served backend/version and reasoning effort unknown)
Original AI-only score: 4.5 / 10 · Scientific importance0-10; overall5of10, criteria4/4/3/3/4 of5
major_revision
First exact250B PDF round, no rescore. Full original opinion plus format-only native clarification retained. All3material objections OPEN. Not acceptance.
Backend identity: unknown. Involved shared-account inherited context; predecessor score visible, not blind or demonstrated independent/human review. Exact served backend/effortunknown. Later249/251-253 excluded fromPDF/proof repair/score.
Report captured · Exact PDF SHA-256 381da284637d6abcd2922ff649550a45243c2000fe16cc5efc65336f5765df2d
- assessment250B-first-round/primary-reading-notes.txt · SHA-256
de74da8a0285082d4a3580a4111e7e8f9fff47816b7a07ee8cd3372ebb8e7d4a - assessment250B-first-round/prompt.txt · SHA-256
b41bd67547e6027f652f0b59e9a77a30acd42a03617b15a9653b68d00b4ab9b8 - assessment250B-first-round/report-hashes.json · SHA-256
7ab1e86b520838743acf3afcd3d698b0bd3da984f1c35909e67125d07767c153 - assessment250B-first-round/response.json · SHA-256
5fa85ab47d42c169eaac45bcfcb8c22ddc76d268011c506887aae2acd5a65393 - assessment250B-first-round/review-context.json · SHA-256
026e75a2372c5b0c8e5d6b499cda190c9ba9681cd72158d660623452a6f974a0 - assessment250B-first-round/scientific-review.txt · SHA-256
9fcb86d4b85ab8b9ac908bcde754761f88bd8c447a66c0fd3e00faf44b9e8ef1 - assessment250B-native-format/clarification-context.json · SHA-256
761170530b397ae3e27d72eca1aae34048e1ed3ed55f7658d46923c969450738 - assessment250B-native-format/clarification.txt · SHA-256
a813261217bedaeb35296c3aaf7f5dac4a10071baeb67f6b63054164508ee494 - assessment250B-native-format/field-mapping.json · SHA-256
528767b5d72781a52a35360afe72daa1cdac9f757de162aec6ffaae864f726fb - assessment250B-native-format/hashes.json · SHA-256
5bf58c0f52d1d2daacb631fc91ba0f241b4ad83139ca63fd054f1a1d1352c221 - assessment250B-native-format/parent-task.txt · SHA-256
b2ad366af744f027ff3db18d20b9112e50b1030efba928831725e7c9d3a32223 - assessment250B-native-format/response-native.json · SHA-256
d8aa642e78f8090a054f295859763a1436a1d29ec64b35039d5eba2045dff2f1 - Author response and present-artifact context · SHA-256
59cabdf376e51554fa8a9a894c27d70fb7e4efb785d1b048c2598064591902f5
A model assessment is not peer review or a correctness certificate. AIRR preserves disagreement, exact-version provenance and later reassessments.
Editorial disclosure
Exact-hash clean native receipt and actual CC-BY/CC0 Working permission verified. First involved-AI importance4.50/10,5stars,criteria4/4/3/3/4; major_revision and all3material objections OPEN. No new human assent, human signoff or acceptance; see DELEGATED-WORKING-DISPOSITION.json.