{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "hoffmann-et-al-chinchilla-2022",
    "title": "Hoffmann and colleagues: compute-optimal language-model training and Chinchilla",
    "summary": "Estimated a compute-optimal balance of model parameters and training tokens, then tested it with the 70-billion-parameter Chinchilla model trained on substantially more tokens than the earlier Gopher model at a comparable training-compute budget.",
    "boundary": "The equal scaling of model size and data is an empirical result under the paper's transformer, data and compute assumptions, not a universal theorem or proof that Chinchilla dominates every later model.",
    "area": "Language-model scaling",
    "topic": "Artificial intelligence",
    "year": 2022,
    "lane": "HUMAN",
    "result": "EMPIRICAL",
    "package_id": "hoffmann-et-al-chinchilla-2022",
    "target": {
      "record_id": "urn:uuid:2e5dd2ff-0d27-41bc-a4eb-06c248705efa",
      "version": 1,
      "content_hash": "sha256:20b76f867ad156de020990ad95053c83ba442b6bc08a8ccd7de23efce5eedcf8"
    },
    "humans": [
      "Jordan Hoffmann",
      "Sebastian Borgeaud",
      "Arthur Mensch",
      "Elena Buchatskaya",
      "Trevor Cai",
      "Eliza Rutherford",
      "Diego de Las Casas",
      "Lisa Anne Hendricks",
      "Johannes Welbl",
      "Aidan Clark",
      "Tom Hennigan",
      "Eric Noland",
      "Katie Millican",
      "George van den Driessche",
      "Bogdan Damoc",
      "Aurelia Guy",
      "Simon Osindero",
      "Karen Simonyan",
      "Erich Elsen",
      "Jack W. Rae",
      "Oriol Vinyals",
      "Laurent Sifre"
    ],
    "authors_complete": true,
    "human_role_note": "Complete named paper byline. Human research, source-reported system execution and experimental validation are distinct; individual effort shares are not measured.",
    "agents": [],
    "related": [],
    "subresults": [],
    "scientific_assessment": {
      "source_review": "SOURCE_CURATED",
      "scientific_review": "NOT_REVIEWED_BY_WRS",
      "formal_replay": "NOT_RUN",
      "provider_attestation": "NONE"
    },
    "sources": [
      {
        "id": "arxiv-chinchilla-2203-15556",
        "title": "Training Compute-Optimal Large Language Models",
        "url": "https://arxiv.org/abs/2203.15556v1"
      },
      {
        "id": "deepmind-chinchilla-2022",
        "title": "An empirical analysis of compute-optimal large language model training — Google DeepMind",
        "url": "https://deepmind.google/blog/an-empirical-analysis-of-compute-optimal-large-language-model-training/"
      }
    ],
    "points": "70000",
    "published_grade": {
      "id": "hoffmann-et-al-chinchilla-2022",
      "entry_id": "hoffmann-et-al-chinchilla-2022",
      "band": "A",
      "modifier": null,
      "target": {
        "record_id": "urn:uuid:2e5dd2ff-0d27-41bc-a4eb-06c248705efa",
        "version": 1,
        "content_hash": "sha256:20b76f867ad156de020990ad95053c83ba442b6bc08a8ccd7de23efce5eedcf8"
      },
      "what_changed": "Estimated a compute-optimal balance of model parameters and training tokens, then tested it with the 70-billion-parameter Chinchilla model trained on substantially more tokens than the earlier Gopher model at a comparable training-compute budget.",
      "scope": "Estimated a compute-optimal balance of model parameters and training tokens, then tested it with the 70-billion-parameter Chinchilla model trained on substantially more tokens than the earlier Gopher model at a comparable training-compute budget. Scope limitation: The equal scaling of model size and data is an empirical result under the paper's transformer, data and compute assumptions, not a universal theorem or proof that Chinchilla dominates every later model.",
      "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:2e5dd2ff-0d27-41bc-a4eb-06c248705efa",
        "version": 1,
        "content_hash": "sha256:20b76f867ad156de020990ad95053c83ba442b6bc08a8ccd7de23efce5eedcf8"
      },
      "edition": "ratings-e039106f30dec177"
    }
  },
  "version": {
    "kind": "PUBLIC_RECORD_VIEW",
    "previous": null,
    "published_at": "2026-09-27T00:00:00Z",
    "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": "Jordan Hoffmann",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Sebastian Borgeaud",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Arthur Mensch",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Elena Buchatskaya",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Trevor Cai",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Eliza Rutherford",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Diego de Las Casas",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Lisa Anne Hendricks",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Johannes Welbl",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Aidan Clark",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Tom Hennigan",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Eric Noland",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Katie Millican",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "George van den Driessche",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Bogdan Damoc",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Aurelia Guy",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Simon Osindero",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Karen Simonyan",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Erich Elsen",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Jack W. Rae",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Oriol Vinyals",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Laurent Sifre",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "Training Compute-Optimal Large Language Models",
            "url": "https://arxiv.org/abs/2203.15556v1"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "An empirical analysis of compute-optimal large language model training — Google DeepMind",
            "url": "https://deepmind.google/blog/an-empirical-analysis-of-compute-optimal-large-language-model-training/"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "The equal scaling of model size and data is an empirical result under the paper's transformer, data and compute assumptions, not a universal theorem or proof that Chinchilla dominates every later model.",
        "domain": "Language-model scaling"
      },
      "sources": [
        {
          "citation": "Training Compute-Optimal Large Language Models",
          "url": "https://arxiv.org/abs/2203.15556v1"
        },
        {
          "citation": "An empirical analysis of compute-optimal large language model training — Google DeepMind",
          "url": "https://deepmind.google/blog/an-empirical-analysis-of-compute-optimal-large-language-model-training/"
        }
      ],
      "statement": "# Hoffmann and colleagues: compute-optimal language-model training and Chinchilla\n\nEstimated a compute-optimal balance of model parameters and training tokens, then tested it with the 70-billion-parameter Chinchilla model trained on substantially more tokens than the earlier Gopher model at a comparable training-compute budget.\n\n## Scientific scope\n\nThe equal scaling of model size and data is an empirical result under the paper's transformer, data and compute assumptions, not a universal theorem or proof that Chinchilla dominates every later model.\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\nJordan Hoffmann; Sebastian Borgeaud; Arthur Mensch; Elena Buchatskaya; Trevor Cai; Eliza Rutherford; Diego de Las Casas; Lisa Anne Hendricks; Johannes Welbl; Aidan Clark; Tom Hennigan; Eric Noland; Katie Millican; George van den Driessche; Bogdan Damoc; Aurelia Guy; Simon Osindero; Karen Simonyan; Erich Elsen; Jack W. Rae; Oriol Vinyals; Laurent Sifre\n\nComplete named paper byline. Human research, source-reported system execution and experimental validation are distinct; individual effort shares are not measured.\n\n\n\n## Original sources\n\n- [Training Compute-Optimal Large Language Models](https://arxiv.org/abs/2203.15556v1)\n- [An empirical analysis of compute-optimal large language model training — Google DeepMind](https://deepmind.google/blog/an-empirical-analysis-of-compute-optimal-large-language-model-training/)\n",
      "summary": "Estimated a compute-optimal balance of model parameters and training tokens, then tested it with the 70-billion-parameter Chinchilla model trained on substantially more tokens than the earlier Gopher model at a comparable training-compute budget.",
      "title": "Hoffmann and colleagues: compute-optimal language-model training and Chinchilla"
    },
    "record_id": "urn:uuid:2e5dd2ff-0d27-41bc-a4eb-06c248705efa",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:2e5dd2ff-0d27-41bc-a4eb-06c248705efa",
      "version": 1,
      "content_hash": "sha256:20b76f867ad156de020990ad95053c83ba442b6bc08a8ccd7de23efce5eedcf8"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:9837b9a9e8eb1b5b1e4b42976e8b16196e44f60f65bddd23d4684d57c5d40e6f"
  },
  "target": {
    "record_id": "urn:uuid:2e5dd2ff-0d27-41bc-a4eb-06c248705efa",
    "version": 1,
    "content_hash": "sha256:20b76f867ad156de020990ad95053c83ba442b6bc08a8ccd7de23efce5eedcf8"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "hoffmann-et-al-chinchilla-2022",
    "entry_id": "hoffmann-et-al-chinchilla-2022",
    "revision": 1,
    "title": "Hoffmann and colleagues: compute-optimal language-model training and Chinchilla",
    "context": "Language-model scaling",
    "track": "Human-designed AI research method or resource",
    "band": "A",
    "modifier": null,
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-09-27",
    "evidence_cutoff": "2026-09-27",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Estimated a compute-optimal balance of model parameters and training tokens, then tested it with the 70-billion-parameter Chinchilla model trained on substantially more tokens than the earlier Gopher model at a comparable training-compute budget.",
    "scope": "Estimated a compute-optimal balance of model parameters and training tokens, then tested it with the 70-billion-parameter Chinchilla model trained on substantially more tokens than the earlier Gopher model at a comparable training-compute budget. Scope limitation: The equal scaling of model size and data is an empirical result under the paper's transformer, data and compute assumptions, not a universal theorem or proof that Chinchilla dominates every later model.",
    "evidence_summary": "The retained primary-source abstract, bibliographic record and stated review scope support this provisional assessment. No independent experimental replication by WRS is claimed. The equal scaling of model size and data is an empirical result under the paper's transformer, data and compute assumptions, not a universal theorem or proof that Chinchilla dominates every later model.",
    "references": [
      {
        "title": "Training Compute-Optimal Large Language Models",
        "url": "https://arxiv.org/abs/2203.15556v1",
        "locator": "Abstract and complete bibliographic byline for 2203.15556v1; not an independent proof or experimental replication."
      },
      {
        "title": "An empirical analysis of compute-optimal large language model training — Google DeepMind",
        "url": "https://deepmind.google/blog/an-empirical-analysis-of-compute-optimal-large-language-model-training/",
        "locator": "Source-linked contribution and stated scientific limits; not an independent replication."
      }
    ],
    "roles": [
      {
        "who": "The named authors listed in the linked research record",
        "credit": "SCIENCE",
        "did": "Complete source byline retained in the research record; human research direction, method development and evaluation. Individual labor shares are not measured."
      }
    ],
    "target": {
      "record_id": "urn:uuid:2e5dd2ff-0d27-41bc-a4eb-06c248705efa",
      "version": 1,
      "content_hash": "sha256:20b76f867ad156de020990ad95053c83ba442b6bc08a8ccd7de23efce5eedcf8"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "70000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}