{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "hu-ruder-et-al-xtreme-family-2020-2021",
    "title": "XTREME and XTREME-R: multilingual transfer evaluation and diagnostics",
    "summary": "Combined multiple classification, structured-prediction, question-answering and retrieval tasks across languages; the revision added harder multilingual-pool retrieval, broader language coverage and behavioral diagnostics.",
    "boundary": "Task and language coverage is uneven, not a full Cartesian product. Constituent datasets keep their original creators. XTREME-R modifies fine-tuning/evaluation for some retained tasks and is not directly score-equivalent to XTREME. Leaderboards compare models plus differing training data, not architecture alone.",
    "area": "Cross-lingual evaluation",
    "topic": "Artificial intelligence",
    "year": 2020,
    "lane": "HUMAN",
    "result": "BENCHMARK",
    "contribution_kind": "EVALUATION_RESOURCE_DESIGN",
    "package_id": "hu-ruder-et-al-xtreme-family-2020-2021",
    "target": {
      "record_id": "urn:uuid:d6c421c1-e11f-4427-a44f-810ca1a22d75",
      "version": 1,
      "content_hash": "sha256:beab63fe17dbcdbaca83876121efd30d38917700a3e6a40426d1de5f4e55cbb7"
    },
    "humans": [
      "Junjie Hu",
      "Sebastian Ruder",
      "Aditya Siddhant",
      "Graham Neubig",
      "Orhan Firat",
      "Melvin Johnson",
      "Noah Constant",
      "Jan Botha",
      "Jinlan Fu",
      "Pengfei Liu"
    ],
    "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-2003-11080v1",
        "title": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
        "url": "https://arxiv.org/abs/2003.11080v1"
      },
      {
        "id": "cur20261015-arxiv-2104-07412v1",
        "title": "XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation",
        "url": "https://arxiv.org/abs/2104.07412v1"
      }
    ],
    "points": "130000",
    "published_grade": {
      "id": "hu-ruder-et-al-xtreme-family-2020-2021",
      "entry_id": "hu-ruder-et-al-xtreme-family-2020-2021",
      "band": "A",
      "modifier": null,
      "target": {
        "record_id": "urn:uuid:d6c421c1-e11f-4427-a44f-810ca1a22d75",
        "version": 1,
        "content_hash": "sha256:beab63fe17dbcdbaca83876121efd30d38917700a3e6a40426d1de5f4e55cbb7"
      },
      "what_changed": "Combined multiple classification, structured-prediction, question-answering and retrieval tasks across languages; the revision added harder multilingual-pool retrieval, broader language coverage and behavioral diagnostics.",
      "scope": "Combined multiple classification, structured-prediction, question-answering and retrieval tasks across languages; the revision added harder multilingual-pool retrieval, broader language coverage and behavioral diagnostics. Task and language coverage is uneven, not a full Cartesian product. Constituent datasets keep their original creators. XTREME-R modifies fine-tuning/evaluation for some retained tasks and is not directly score-equivalent to XTREME. Leaderboards compare models plus differing training data, not architecture alone.",
      "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:d6c421c1-e11f-4427-a44f-810ca1a22d75",
        "version": 1,
        "content_hash": "sha256:beab63fe17dbcdbaca83876121efd30d38917700a3e6a40426d1de5f4e55cbb7"
      },
      "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": "Junjie Hu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
            "url": "https://arxiv.org/abs/2003.11080v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Sebastian Ruder",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
            "url": "https://arxiv.org/abs/2003.11080v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Aditya Siddhant",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
            "url": "https://arxiv.org/abs/2003.11080v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Graham Neubig",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
            "url": "https://arxiv.org/abs/2003.11080v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Orhan Firat",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
            "url": "https://arxiv.org/abs/2003.11080v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Melvin Johnson",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
            "url": "https://arxiv.org/abs/2003.11080v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Noah Constant",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation",
            "url": "https://arxiv.org/abs/2104.07412v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Jan Botha",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation",
            "url": "https://arxiv.org/abs/2104.07412v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Jinlan Fu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation",
            "url": "https://arxiv.org/abs/2104.07412v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Pengfei Liu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation",
            "url": "https://arxiv.org/abs/2104.07412v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
            "url": "https://arxiv.org/abs/2003.11080v1"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation",
            "url": "https://arxiv.org/abs/2104.07412v1"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "Task and language coverage is uneven, not a full Cartesian product. Constituent datasets keep their original creators. XTREME-R modifies fine-tuning/evaluation for some retained tasks and is not directly score-equivalent to XTREME. Leaderboards compare models plus differing training data, not architecture alone.",
        "domain": "Cross-lingual evaluation"
      },
      "sources": [
        {
          "citation": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
          "url": "https://arxiv.org/abs/2003.11080v1"
        },
        {
          "citation": "XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation",
          "url": "https://arxiv.org/abs/2104.07412v1"
        }
      ],
      "statement": "# XTREME and XTREME-R: multilingual transfer evaluation and diagnostics\n\nCombined multiple classification, structured-prediction, question-answering and retrieval tasks across languages; the revision added harder multilingual-pool retrieval, broader language coverage and behavioral diagnostics.\n\n## Scientific scope\n\nTask and language coverage is uneven, not a full Cartesian product. Constituent datasets keep their original creators. XTREME-R modifies fine-tuning/evaluation for some retained tasks and is not directly score-equivalent to XTREME. Leaderboards compare models plus differing training data, not architecture alone.\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\nJunjie Hu; Sebastian Ruder; Aditya Siddhant; Graham Neubig; Orhan Firat; Melvin Johnson; Noah Constant; Jan Botha; Jinlan Fu; Pengfei Liu\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- [XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization](https://arxiv.org/abs/2003.11080v1)\n- [XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation](https://arxiv.org/abs/2104.07412v1)\n",
      "summary": "Combined multiple classification, structured-prediction, question-answering and retrieval tasks across languages; the revision added harder multilingual-pool retrieval, broader language coverage and behavioral diagnostics.",
      "title": "XTREME and XTREME-R: multilingual transfer evaluation and diagnostics"
    },
    "record_id": "urn:uuid:d6c421c1-e11f-4427-a44f-810ca1a22d75",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:d6c421c1-e11f-4427-a44f-810ca1a22d75",
      "version": 1,
      "content_hash": "sha256:beab63fe17dbcdbaca83876121efd30d38917700a3e6a40426d1de5f4e55cbb7"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:44dc0b3a6df3acdfe67550b266661e906841d8d1b189025773e86c175a104bd6"
  },
  "target": {
    "record_id": "urn:uuid:d6c421c1-e11f-4427-a44f-810ca1a22d75",
    "version": 1,
    "content_hash": "sha256:beab63fe17dbcdbaca83876121efd30d38917700a3e6a40426d1de5f4e55cbb7"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "hu-ruder-et-al-xtreme-family-2020-2021",
    "entry_id": "hu-ruder-et-al-xtreme-family-2020-2021",
    "revision": 1,
    "title": "XTREME and XTREME-R: multilingual transfer evaluation and diagnostics",
    "context": "Cross-lingual evaluation",
    "track": "Evaluation resource design",
    "band": "A",
    "modifier": null,
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-10-16",
    "evidence_cutoff": "2026-10-16",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Combined multiple classification, structured-prediction, question-answering and retrieval tasks across languages; the revision added harder multilingual-pool retrieval, broader language coverage and behavioral diagnostics.",
    "scope": "Combined multiple classification, structured-prediction, question-answering and retrieval tasks across languages; the revision added harder multilingual-pool retrieval, broader language coverage and behavioral diagnostics. Task and language coverage is uneven, not a full Cartesian product. Constituent datasets keep their original creators. XTREME-R modifies fine-tuning/evaluation for some retained tasks and is not directly score-equivalent to XTREME. Leaderboards compare models plus differing training data, not architecture alone.",
    "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. Task and language coverage is uneven, not a full Cartesian product. Constituent datasets keep their original creators. XTREME-R modifies fine-tuning/evaluation for some retained tasks and is not directly score-equivalent to XTREME. Leaderboards compare models plus differing training data, not architecture alone.",
    "references": [
      {
        "title": "XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization",
        "url": "https://arxiv.org/abs/2003.11080v1",
        "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification."
      },
      {
        "title": "XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluation",
        "url": "https://arxiv.org/abs/2104.07412v1",
        "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:d6c421c1-e11f-4427-a44f-810ca1a22d75",
      "version": 1,
      "content_hash": "sha256:beab63fe17dbcdbaca83876121efd30d38917700a3e6a40426d1de5f4e55cbb7"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "130000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}