{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "huang-et-al-layoutlmv3-unified-masking-2022",
    "title": "Huang et al.: LayoutLMv3 unified text-image masking and alignment",
    "summary": "Unified masked text and image modeling and introduced word-patch alignment, using a common document Transformer representation for text- and image-centered tasks.",
    "boundary": "Extends the LayoutLM family through objective/representation design, not a new contribution for each dataset or model checkpoint. Uses inherited masked modeling and patch/tokenization ideas; no universal document comprehension guarantee.",
    "area": "Multimodal document representation learning",
    "topic": "Artificial intelligence",
    "year": 2022,
    "lane": "HUMAN",
    "result": "EMPIRICAL",
    "package_id": "huang-et-al-layoutlmv3-unified-masking-2022",
    "target": {
      "record_id": "urn:uuid:79abfb72-671a-4e07-a368-595f894b0ac4",
      "version": 1,
      "content_hash": "sha256:eb9a022deb6e2dac728ef76be3d050677e92224a917b71c96281bec8c96c96e5"
    },
    "humans": [
      "Yupan Huang",
      "Tengchao Lv",
      "Lei Cui",
      "Yutong Lu",
      "Furu Wei"
    ],
    "authors_complete": true,
    "human_role_note": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.",
    "agents": [],
    "related": [],
    "subresults": [],
    "scientific_assessment": {
      "source_review": "SOURCE_CURATED",
      "scientific_review": "NOT_REVIEWED_BY_WRS",
      "formal_replay": "NOT_RUN",
      "provider_attestation": "NONE"
    },
    "sources": [
      {
        "id": "cur20261015-arxiv-2204-08387v1",
        "title": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
        "url": "https://arxiv.org/abs/2204.08387v1"
      },
      {
        "id": "cur20261015-support-huang-et-al-layoutlmv3-unified-masking-2022",
        "title": "Huang et al.: LayoutLMv3 unified text-image masking and alignment — official implementation",
        "url": "https://github.com/microsoft/unilm/tree/master/layoutlmv3"
      }
    ],
    "points": "75000",
    "published_grade": {
      "id": "huang-et-al-layoutlmv3-unified-masking-2022",
      "entry_id": "huang-et-al-layoutlmv3-unified-masking-2022",
      "band": "B",
      "modifier": "+",
      "target": {
        "record_id": "urn:uuid:79abfb72-671a-4e07-a368-595f894b0ac4",
        "version": 1,
        "content_hash": "sha256:eb9a022deb6e2dac728ef76be3d050677e92224a917b71c96281bec8c96c96e5"
      },
      "what_changed": "Unified masked text and image modeling and introduced word-patch alignment, using a common document Transformer representation for text- and image-centered tasks.",
      "scope": "Unified masked text and image modeling and introduced word-patch alignment, using a common document Transformer representation for text- and image-centered tasks. Extends the LayoutLM family through objective/representation design, not a new contribution for each dataset or model checkpoint. Uses inherited masked modeling and patch/tokenization ideas; no universal document comprehension guarantee.",
      "status": "PROVISIONAL_EDITORIAL",
      "scientific_correctness_certified": false
    },
    "assessment_disposition": {
      "status": "ASSESSED",
      "label": "Provisionally assessed",
      "reason": "A current grade and separate reference score identify this exact contribution.",
      "target": {
        "record_id": "urn:uuid:79abfb72-671a-4e07-a368-595f894b0ac4",
        "version": 1,
        "content_hash": "sha256:eb9a022deb6e2dac728ef76be3d050677e92224a917b71c96281bec8c96c96e5"
      },
      "edition": "ratings-e039106f30dec177"
    }
  },
  "version": {
    "kind": "PUBLIC_RECORD_VIEW",
    "previous": null,
    "published_at": "2026-10-16T00:00:00Z",
    "publisher": {
      "principal_id": "urn:uuid:80116931-785b-40a2-a24b-9d59db0c7e33",
      "role": "WRS_EDITORIAL_PUBLICATION"
    },
    "record": {
      "contributors": [
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Yupan Huang",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "discoverer"
          ],
          "source": {
            "citation": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
            "url": "https://arxiv.org/abs/2204.08387v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Tengchao Lv",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "discoverer"
          ],
          "source": {
            "citation": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
            "url": "https://arxiv.org/abs/2204.08387v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Lei Cui",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "discoverer"
          ],
          "source": {
            "citation": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
            "url": "https://arxiv.org/abs/2204.08387v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Yutong Lu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "discoverer"
          ],
          "source": {
            "citation": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
            "url": "https://arxiv.org/abs/2204.08387v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Furu Wei",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "discoverer"
          ],
          "source": {
            "citation": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
            "url": "https://arxiv.org/abs/2204.08387v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
            "url": "https://arxiv.org/abs/2204.08387v1"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "Huang et al.: LayoutLMv3 unified text-image masking and alignment — official implementation",
            "url": "https://github.com/microsoft/unilm/tree/master/layoutlmv3"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "Extends the LayoutLM family through objective/representation design, not a new contribution for each dataset or model checkpoint. Uses inherited masked modeling and patch/tokenization ideas; no universal document comprehension guarantee.",
        "domain": "Multimodal document representation learning"
      },
      "sources": [
        {
          "citation": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
          "url": "https://arxiv.org/abs/2204.08387v1"
        },
        {
          "citation": "Huang et al.: LayoutLMv3 unified text-image masking and alignment — official implementation",
          "url": "https://github.com/microsoft/unilm/tree/master/layoutlmv3"
        }
      ],
      "statement": "# Huang et al.: LayoutLMv3 unified text-image masking and alignment\n\nUnified masked text and image modeling and introduced word-patch alignment, using a common document Transformer representation for text- and image-centered tasks.\n\n## Scientific scope\n\nExtends the LayoutLM family through objective/representation design, not a new contribution for each dataset or model checkpoint. Uses inherited masked modeling and patch/tokenization ideas; no universal document comprehension guarantee.\n\n## Reading status\n\nThis is a source-linked WRS catalogue note. A full explanatory guide and independent scientific reconstruction have not been completed for this entry. Read the identified original work for its complete narrative and arguments.\n\n## People and systems\n\nYupan Huang; Tengchao Lv; Lei Cui; Yutong Lu; Furu Wei\n\nSource-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.\n\n\n\n## Original sources\n\n- [LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking](https://arxiv.org/abs/2204.08387v1)\n- [Huang et al.: LayoutLMv3 unified text-image masking and alignment — official implementation](https://github.com/microsoft/unilm/tree/master/layoutlmv3)\n",
      "summary": "Unified masked text and image modeling and introduced word-patch alignment, using a common document Transformer representation for text- and image-centered tasks.",
      "title": "Huang et al.: LayoutLMv3 unified text-image masking and alignment"
    },
    "record_id": "urn:uuid:79abfb72-671a-4e07-a368-595f894b0ac4",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:79abfb72-671a-4e07-a368-595f894b0ac4",
      "version": 1,
      "content_hash": "sha256:eb9a022deb6e2dac728ef76be3d050677e92224a917b71c96281bec8c96c96e5"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:38fd0d09999fec152bdf0f079955c3aff2ab66cee9c74ad4059fd15e7455669e"
  },
  "target": {
    "record_id": "urn:uuid:79abfb72-671a-4e07-a368-595f894b0ac4",
    "version": 1,
    "content_hash": "sha256:eb9a022deb6e2dac728ef76be3d050677e92224a917b71c96281bec8c96c96e5"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "huang-et-al-layoutlmv3-unified-masking-2022",
    "entry_id": "huang-et-al-layoutlmv3-unified-masking-2022",
    "revision": 1,
    "title": "Huang et al.: LayoutLMv3 unified text-image masking and alignment",
    "context": "Multimodal document representation learning",
    "track": "Human-designed research method or resource",
    "band": "B",
    "modifier": "+",
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-10-16",
    "evidence_cutoff": "2026-10-16",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Unified masked text and image modeling and introduced word-patch alignment, using a common document Transformer representation for text- and image-centered tasks.",
    "scope": "Unified masked text and image modeling and introduced word-patch alignment, using a common document Transformer representation for text- and image-centered tasks. Extends the LayoutLM family through objective/representation design, not a new contribution for each dataset or model checkpoint. Uses inherited masked modeling and patch/tokenization ideas; no universal document comprehension guarantee.",
    "evidence_summary": "The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Extends the LayoutLM family through objective/representation design, not a new contribution for each dataset or model checkpoint. Uses inherited masked modeling and patch/tokenization ideas; no universal document comprehension guarantee.",
    "references": [
      {
        "title": "LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking",
        "url": "https://arxiv.org/abs/2204.08387v1",
        "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification."
      },
      {
        "title": "Huang et al.: LayoutLMv3 unified text-image masking and alignment — official implementation",
        "url": "https://github.com/microsoft/unilm/tree/master/layoutlmv3",
        "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification."
      }
    ],
    "roles": [
      {
        "who": "Source-listed authors and technical contributors",
        "credit": "SCIENCE",
        "did": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities."
      }
    ],
    "target": {
      "record_id": "urn:uuid:79abfb72-671a-4e07-a368-595f894b0ac4",
      "version": 1,
      "content_hash": "sha256:eb9a022deb6e2dac728ef76be3d050677e92224a917b71c96281bec8c96c96e5"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "75000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}