[
  {
    "id": "ARR-2026-7EPM3V3DP796BBRR-v15",
    "type": "article",
    "title": "Rectangular observations for finite matrix gradients: observation classes and critical endpoints",
    "author": [
      {
        "literal": "Lluis Eriksson"
      }
    ],
    "issued": {
      "date-parts": [
        [
          2026,
          11,
          11
        ]
      ]
    },
    "publisher": "AIRR.SCIENCE",
    "archive": "AIRR.SCIENCE",
    "archive_location": "ARR-2026-7EPM3V3DP796BBRR",
    "version": "v15",
    "genre": "Research preprint",
    "URL": "https://airr.science/papers/ARR-2026-7EPM3V3DP796BBRR/versions/v15/",
    "note": "Research preprint; ARR-2026-7EPM3V3DP796BBRR v15. Not peer reviewed by AIRR. Working paper; not admitted to the AIRR accepted collection.",
    "abstract": "For finite periodic lattices and symmetric matrices acting on actual forward-gradient stacks, we give a coefficient-only rectangular observation construction with capped-budget dimension control and a finite profile-dependent quadratic penalty. A scalar balanced-ring proof and a weak critical endpoint analysis are included. For common full-vector partition means on the actual flat coexact domain we prove the sharp dimension-only lower bound g(D-1)(R-1). Component-adapted rectangles admit an additive bound for coefficients block diagonal by actual field component; that hypothesis is retained. On one positive constant internally coupled family with D>=3 and growing g=n^D, any map separating the specified internal channels has actual coexact rank Omega_D(n^(2D-2)), and real spatial modes give the matching upper order. These rows may be nonlocal. Mixed internal observations escape the obstruction on the same profile. Historical priority, practical usefulness and conditioning remain unestablished. No interacting/covariant producer, continuum theorem or physical gap follows. This exact30-page candidate has one first-round AIassessment4.10/10 and full original reports, clarification, author response and reproduction archive; current native-format recommendation is major_revision with all three material objections open. It is proposed as a working paper, not editorial acceptance."
  }
]
