{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "xiao-chen-et-al-bge-resource-hybrid-family-2023",
    "title": "BGE/C-Pack and M3: embedding resources and self-distilled hybrid retrieval",
    "summary": "Packaged Chinese evaluation/training resources and a staged embedding recipe, then extended the family with shared dense, lexical and multi-vector retrieval heads whose integrated scores teach each head through self-distillation.",
    "boundary": "C-Pack original C-TEM resource names and later BGE-M3 extension are source-specific stages, not retroactively identical models. M3's integrated teacher is an empirical training choice, not guaranteed more accurate than each head. Dense, sparse and late-interaction components are inherited; full multi-vector scoring has extra cost. The retrieval method does not itself establish a medical benefit.",
    "area": "Multilingual text retrieval",
    "topic": "Artificial intelligence",
    "year": 2023,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "xiao-chen-et-al-bge-resource-hybrid-family-2023",
    "target": {
      "record_id": "urn:uuid:46de459c-4827-4ae4-a11f-8f97433d7cba",
      "version": 1,
      "content_hash": "sha256:33f88329fc2773aa128b7593f801095266b85564c3042ebc2b03484983803847"
    },
    "humans": [
      "Shitao Xiao",
      "Zheng Liu",
      "Peitian Zhang",
      "Niklas Muennighoff",
      "Jianlv Chen",
      "Kun Luo",
      "Defu Lian"
    ],
    "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-2309-07597v1",
        "title": "C-Pack: Packaged Resources To Advance General Chinese Embedding",
        "url": "https://arxiv.org/abs/2309.07597v1"
      },
      {
        "id": "cur20261015-arxiv-2402-03216v1",
        "title": "BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation",
        "url": "https://arxiv.org/abs/2402.03216v1"
      }
    ],
    "points": "120000",
    "published_grade": {
      "id": "xiao-chen-et-al-bge-resource-hybrid-family-2023",
      "entry_id": "xiao-chen-et-al-bge-resource-hybrid-family-2023",
      "band": "A",
      "modifier": null,
      "target": {
        "record_id": "urn:uuid:46de459c-4827-4ae4-a11f-8f97433d7cba",
        "version": 1,
        "content_hash": "sha256:33f88329fc2773aa128b7593f801095266b85564c3042ebc2b03484983803847"
      },
      "what_changed": "Packaged Chinese evaluation/training resources and a staged embedding recipe, then extended the family with shared dense, lexical and multi-vector retrieval heads whose integrated scores teach each head through self-distillation.",
      "scope": "Packaged Chinese evaluation/training resources and a staged embedding recipe, then extended the family with shared dense, lexical and multi-vector retrieval heads whose integrated scores teach each head through self-distillation. C-Pack original C-TEM resource names and later BGE-M3 extension are source-specific stages, not retroactively identical models. M3's integrated teacher is an empirical training choice, not guaranteed more accurate than each head. Dense, sparse and late-interaction components are inherited; full multi-vector scoring has extra cost. The retrieval method does not itself establish a medical benefit.",
      "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:46de459c-4827-4ae4-a11f-8f97433d7cba",
        "version": 1,
        "content_hash": "sha256:33f88329fc2773aa128b7593f801095266b85564c3042ebc2b03484983803847"
      },
      "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": "Shitao Xiao",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "C-Pack: Packaged Resources To Advance General Chinese Embedding",
            "url": "https://arxiv.org/abs/2309.07597v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Zheng Liu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "C-Pack: Packaged Resources To Advance General Chinese Embedding",
            "url": "https://arxiv.org/abs/2309.07597v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Peitian Zhang",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "C-Pack: Packaged Resources To Advance General Chinese Embedding",
            "url": "https://arxiv.org/abs/2309.07597v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Niklas Muennighoff",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "C-Pack: Packaged Resources To Advance General Chinese Embedding",
            "url": "https://arxiv.org/abs/2309.07597v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Jianlv Chen",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation",
            "url": "https://arxiv.org/abs/2402.03216v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Kun Luo",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation",
            "url": "https://arxiv.org/abs/2402.03216v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Defu Lian",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation",
            "url": "https://arxiv.org/abs/2402.03216v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "C-Pack: Packaged Resources To Advance General Chinese Embedding",
            "url": "https://arxiv.org/abs/2309.07597v1"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation",
            "url": "https://arxiv.org/abs/2402.03216v1"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "C-Pack original C-TEM resource names and later BGE-M3 extension are source-specific stages, not retroactively identical models. M3's integrated teacher is an empirical training choice, not guaranteed more accurate than each head. Dense, sparse and late-interaction components are inherited; full multi-vector scoring has extra cost. The retrieval method does not itself establish a medical benefit.",
        "domain": "Multilingual text retrieval"
      },
      "sources": [
        {
          "citation": "C-Pack: Packaged Resources To Advance General Chinese Embedding",
          "url": "https://arxiv.org/abs/2309.07597v1"
        },
        {
          "citation": "BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation",
          "url": "https://arxiv.org/abs/2402.03216v1"
        }
      ],
      "statement": "# BGE/C-Pack and M3: embedding resources and self-distilled hybrid retrieval\n\nPackaged Chinese evaluation/training resources and a staged embedding recipe, then extended the family with shared dense, lexical and multi-vector retrieval heads whose integrated scores teach each head through self-distillation.\n\n## Scientific scope\n\nC-Pack original C-TEM resource names and later BGE-M3 extension are source-specific stages, not retroactively identical models. M3's integrated teacher is an empirical training choice, not guaranteed more accurate than each head. Dense, sparse and late-interaction components are inherited; full multi-vector scoring has extra cost. The retrieval method does not itself establish a medical benefit.\n\n## People and systems\n\nShitao Xiao; Zheng Liu; Peitian Zhang; Niklas Muennighoff; Jianlv Chen; Kun Luo; Defu Lian\n\n\n\n## Original sources\n\n- [C-Pack: Packaged Resources To Advance General Chinese Embedding](https://arxiv.org/abs/2309.07597v1)\n- [BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation](https://arxiv.org/abs/2402.03216v1)\n",
      "summary": "Packaged Chinese evaluation/training resources and a staged embedding recipe, then extended the family with shared dense, lexical and multi-vector retrieval heads whose integrated scores teach each head through self-distillation.",
      "title": "BGE/C-Pack and M3: embedding resources and self-distilled hybrid retrieval"
    },
    "record_id": "urn:uuid:46de459c-4827-4ae4-a11f-8f97433d7cba",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:46de459c-4827-4ae4-a11f-8f97433d7cba",
      "version": 1,
      "content_hash": "sha256:33f88329fc2773aa128b7593f801095266b85564c3042ebc2b03484983803847"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:d95550d858a7fab01fa41ae048b0e068c6a1efc00f0ef63d9ed14cbe35897fed"
  },
  "target": {
    "record_id": "urn:uuid:46de459c-4827-4ae4-a11f-8f97433d7cba",
    "version": 1,
    "content_hash": "sha256:33f88329fc2773aa128b7593f801095266b85564c3042ebc2b03484983803847"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "xiao-chen-et-al-bge-resource-hybrid-family-2023",
    "entry_id": "xiao-chen-et-al-bge-resource-hybrid-family-2023",
    "revision": 1,
    "title": "BGE/C-Pack and M3: embedding resources and self-distilled hybrid retrieval",
    "context": "Multilingual text retrieval",
    "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": "Packaged Chinese evaluation/training resources and a staged embedding recipe, then extended the family with shared dense, lexical and multi-vector retrieval heads whose integrated scores teach each head through self-distillation.",
    "scope": "Packaged Chinese evaluation/training resources and a staged embedding recipe, then extended the family with shared dense, lexical and multi-vector retrieval heads whose integrated scores teach each head through self-distillation. C-Pack original C-TEM resource names and later BGE-M3 extension are source-specific stages, not retroactively identical models. M3's integrated teacher is an empirical training choice, not guaranteed more accurate than each head. Dense, sparse and late-interaction components are inherited; full multi-vector scoring has extra cost. The retrieval method does not itself establish a medical benefit.",
    "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. C-Pack original C-TEM resource names and later BGE-M3 extension are source-specific stages, not retroactively identical models. M3's integrated teacher is an empirical training choice, not guaranteed more accurate than each head. Dense, sparse and late-interaction components are inherited; full multi-vector scoring has extra cost. The retrieval method does not itself establish a medical benefit.",
    "references": [
      {
        "title": "C-Pack: Packaged Resources To Advance General Chinese Embedding",
        "url": "https://arxiv.org/abs/2309.07597v1",
        "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification."
      },
      {
        "title": "BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation",
        "url": "https://arxiv.org/abs/2402.03216v1",
        "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:46de459c-4827-4ae4-a11f-8f97433d7cba",
      "version": 1,
      "content_hash": "sha256:33f88329fc2773aa128b7593f801095266b85564c3042ebc2b03484983803847"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "120000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}