{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "baevski-et-al-data2vec-contextual-targets-2022",
    "title": "data2vec family: contextual targets and efficient masked self-distillation",
    "summary": "Regressed contextual latent representations from an EMA teacher's full input using masked student views, applying one self-distillation objective across separate speech, image and text settings. data2vec2 skipped masked inputs, used a convolutional decoder and amortized teacher computation across several masked student views.",
    "boundary": "Inherits EMA teachers, BYOL/DINO-style self-distillation and masking. Same learning method across modalities is not a single jointly trained multimodal model: feature encoders and masking remain modality specific. Not invention of all latent prediction or evidence for a biological learning theory; benchmark gains and anti-collapse behavior have configuration limits. Version2 inherits visible-only encoding and multi-mask reuse; reported16.4fold image pretraining reduction is protocol-specific, not every modality or end-to-end workflow.",
    "area": "Self-supervised representation learning",
    "topic": "Artificial intelligence",
    "year": 2022,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "baevski-et-al-data2vec-contextual-targets-2022",
    "target": {
      "record_id": "urn:uuid:cc54bead-561c-4365-a272-69e95921da35",
      "version": 1,
      "content_hash": "sha256:e86a052edb31f5690b9526cd7082a9175f99ea1effd67d18ca4dd1c4061d476f"
    },
    "humans": [
      "Alexei Baevski",
      "Wei-Ning Hsu",
      "Qiantong Xu",
      "Arun Babu",
      "Jiatao Gu",
      "Michael Auli"
    ],
    "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-2202-03555v1",
        "title": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
        "url": "https://arxiv.org/abs/2202.03555v1"
      },
      {
        "id": "cur20261015-arxiv-2212-07525v1",
        "title": "Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language",
        "url": "https://arxiv.org/abs/2212.07525v1"
      }
    ],
    "points": "125000",
    "published_grade": {
      "id": "baevski-et-al-data2vec-contextual-targets-2022",
      "entry_id": "baevski-et-al-data2vec-contextual-targets-2022",
      "band": "A",
      "modifier": null,
      "target": {
        "record_id": "urn:uuid:cc54bead-561c-4365-a272-69e95921da35",
        "version": 1,
        "content_hash": "sha256:e86a052edb31f5690b9526cd7082a9175f99ea1effd67d18ca4dd1c4061d476f"
      },
      "what_changed": "Regressed contextual latent representations from an EMA teacher's full input using masked student views, applying one self-distillation objective across separate speech, image and text settings. data2vec2 skipped masked inputs, used a convolutional decoder and amortized teacher computation across several masked student views.",
      "scope": "Regressed contextual latent representations from an EMA teacher's full input using masked student views, applying one self-distillation objective across separate speech, image and text settings. data2vec2 skipped masked inputs, used a convolutional decoder and amortized teacher computation across several masked student views. Inherits EMA teachers, BYOL/DINO-style self-distillation and masking. Same learning method across modalities is not a single jointly trained multimodal model: feature encoders and masking remain modality specific. Not invention of all latent prediction or evidence for a biological learning theory; benchmark gains and anti-collapse behavior have configuration limits. Version2 inherits visible-only encoding and multi-mask reuse; reported16.4fold image pretraining reduction is protocol-specific, not every modality or end-to-end workflow.",
      "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:cc54bead-561c-4365-a272-69e95921da35",
        "version": 1,
        "content_hash": "sha256:e86a052edb31f5690b9526cd7082a9175f99ea1effd67d18ca4dd1c4061d476f"
      },
      "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": "Alexei Baevski",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
            "url": "https://arxiv.org/abs/2202.03555v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Wei-Ning Hsu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
            "url": "https://arxiv.org/abs/2202.03555v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Qiantong Xu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
            "url": "https://arxiv.org/abs/2202.03555v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Arun Babu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
            "url": "https://arxiv.org/abs/2202.03555v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Jiatao Gu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
            "url": "https://arxiv.org/abs/2202.03555v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Michael Auli",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
            "url": "https://arxiv.org/abs/2202.03555v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
            "url": "https://arxiv.org/abs/2202.03555v1"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language",
            "url": "https://arxiv.org/abs/2212.07525v1"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "Inherits EMA teachers, BYOL/DINO-style self-distillation and masking. Same learning method across modalities is not a single jointly trained multimodal model: feature encoders and masking remain modality specific. Not invention of all latent prediction or evidence for a biological learning theory; benchmark gains and anti-collapse behavior have configuration limits. Version2 inherits visible-only encoding and multi-mask reuse; reported16.4fold image pretraining reduction is protocol-specific, not every modality or end-to-end workflow.",
        "domain": "Self-supervised representation learning"
      },
      "sources": [
        {
          "citation": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
          "url": "https://arxiv.org/abs/2202.03555v1"
        },
        {
          "citation": "Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language",
          "url": "https://arxiv.org/abs/2212.07525v1"
        }
      ],
      "statement": "# data2vec family: contextual targets and efficient masked self-distillation\n\nRegressed contextual latent representations from an EMA teacher's full input using masked student views, applying one self-distillation objective across separate speech, image and text settings. data2vec2 skipped masked inputs, used a convolutional decoder and amortized teacher computation across several masked student views.\n\n## Scientific scope\n\nInherits EMA teachers, BYOL/DINO-style self-distillation and masking. Same learning method across modalities is not a single jointly trained multimodal model: feature encoders and masking remain modality specific. Not invention of all latent prediction or evidence for a biological learning theory; benchmark gains and anti-collapse behavior have configuration limits. Version2 inherits visible-only encoding and multi-mask reuse; reported16.4fold image pretraining reduction is protocol-specific, not every modality or end-to-end workflow.\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\nAlexei Baevski; Wei-Ning Hsu; Qiantong Xu; Arun Babu; Jiatao Gu; Michael Auli\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- [data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language](https://arxiv.org/abs/2202.03555v1)\n- [Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language](https://arxiv.org/abs/2212.07525v1)\n",
      "summary": "Regressed contextual latent representations from an EMA teacher's full input using masked student views, applying one self-distillation objective across separate speech, image and text settings. data2vec2 skipped masked inputs, used a convolutional decoder and amortized teacher computation across several masked student views.",
      "title": "data2vec family: contextual targets and efficient masked self-distillation"
    },
    "record_id": "urn:uuid:cc54bead-561c-4365-a272-69e95921da35",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:cc54bead-561c-4365-a272-69e95921da35",
      "version": 1,
      "content_hash": "sha256:e86a052edb31f5690b9526cd7082a9175f99ea1effd67d18ca4dd1c4061d476f"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:b0cfeb8c1208ccf82ebbd4b065e324cee5cac06a480579eb6c15bbe571cfcba5"
  },
  "target": {
    "record_id": "urn:uuid:cc54bead-561c-4365-a272-69e95921da35",
    "version": 1,
    "content_hash": "sha256:e86a052edb31f5690b9526cd7082a9175f99ea1effd67d18ca4dd1c4061d476f"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "baevski-et-al-data2vec-contextual-targets-2022",
    "entry_id": "baevski-et-al-data2vec-contextual-targets-2022",
    "revision": 1,
    "title": "data2vec family: contextual targets and efficient masked self-distillation",
    "context": "Self-supervised representation learning",
    "track": "Human-designed research method or resource",
    "band": "A",
    "modifier": null,
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-10-16",
    "evidence_cutoff": "2026-10-16",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Regressed contextual latent representations from an EMA teacher's full input using masked student views, applying one self-distillation objective across separate speech, image and text settings. data2vec2 skipped masked inputs, used a convolutional decoder and amortized teacher computation across several masked student views.",
    "scope": "Regressed contextual latent representations from an EMA teacher's full input using masked student views, applying one self-distillation objective across separate speech, image and text settings. data2vec2 skipped masked inputs, used a convolutional decoder and amortized teacher computation across several masked student views. Inherits EMA teachers, BYOL/DINO-style self-distillation and masking. Same learning method across modalities is not a single jointly trained multimodal model: feature encoders and masking remain modality specific. Not invention of all latent prediction or evidence for a biological learning theory; benchmark gains and anti-collapse behavior have configuration limits. Version2 inherits visible-only encoding and multi-mask reuse; reported16.4fold image pretraining reduction is protocol-specific, not every modality or end-to-end workflow.",
    "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. Inherits EMA teachers, BYOL/DINO-style self-distillation and masking. Same learning method across modalities is not a single jointly trained multimodal model: feature encoders and masking remain modality specific. Not invention of all latent prediction or evidence for a biological learning theory; benchmark gains and anti-collapse behavior have configuration limits. Version2 inherits visible-only encoding and multi-mask reuse; reported16.4fold image pretraining reduction is protocol-specific, not every modality or end-to-end workflow.",
    "references": [
      {
        "title": "data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language",
        "url": "https://arxiv.org/abs/2202.03555v1",
        "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification."
      },
      {
        "title": "Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language",
        "url": "https://arxiv.org/abs/2212.07525v1",
        "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:cc54bead-561c-4365-a272-69e95921da35",
      "version": 1,
      "content_hash": "sha256:e86a052edb31f5690b9526cd7082a9175f99ea1effd67d18ca4dd1c4061d476f"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "125000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}