{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "barham-et-al-pathways-asynchronous-dataflow-2022",
    "title": "Pathways: asynchronous distributed accelerator dataflow",
    "summary": "Used sharded asynchronous operators over futures and gang scheduling to decouple control-plane execution from dependent accelerator work under a single-controller programming model.",
    "boundary": "The orchestration-system contribution is distinct from PaLM's language-model training. Inherits dataflow/SPMD/MPMD systems; measured utilization on selected TPU workloads is not universal100percent utilization, failure freedom or proof of optimal resource scheduling.",
    "area": "Large-scale AI orchestration systems",
    "topic": "Artificial intelligence",
    "year": 2022,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "barham-et-al-pathways-asynchronous-dataflow-2022",
    "target": {
      "record_id": "urn:uuid:b84cb4f0-821d-46e1-a640-7e020318e456",
      "version": 1,
      "content_hash": "sha256:c748075d185006cd08fcb7a91757d1661052f01f0bcb366c092b10c261594660"
    },
    "humans": [
      "Paul Barham",
      "Aakanksha Chowdhery",
      "Jeff Dean",
      "Sanjay Ghemawat",
      "Steven Hand",
      "Dan Hurt",
      "Michael Isard",
      "Hyeontaek Lim",
      "Ruoming Pang",
      "Sudip Roy",
      "Brennan Saeta",
      "Parker Schuh",
      "Ryan Sepassi",
      "Laurent El Shafey",
      "Chandramohan A. Thekkath",
      "Yonghui Wu"
    ],
    "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-2203-12533v1",
        "title": "Pathways: Asynchronous Distributed Dataflow for ML",
        "url": "https://arxiv.org/abs/2203.12533v1"
      }
    ],
    "points": "130000",
    "published_grade": {
      "id": "barham-et-al-pathways-asynchronous-dataflow-2022",
      "entry_id": "barham-et-al-pathways-asynchronous-dataflow-2022",
      "band": "A",
      "modifier": null,
      "target": {
        "record_id": "urn:uuid:b84cb4f0-821d-46e1-a640-7e020318e456",
        "version": 1,
        "content_hash": "sha256:c748075d185006cd08fcb7a91757d1661052f01f0bcb366c092b10c261594660"
      },
      "what_changed": "Used sharded asynchronous operators over futures and gang scheduling to decouple control-plane execution from dependent accelerator work under a single-controller programming model.",
      "scope": "Used sharded asynchronous operators over futures and gang scheduling to decouple control-plane execution from dependent accelerator work under a single-controller programming model. The orchestration-system contribution is distinct from PaLM's language-model training. Inherits dataflow/SPMD/MPMD systems; measured utilization on selected TPU workloads is not universal100percent utilization, failure freedom or proof of optimal resource scheduling.",
      "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:b84cb4f0-821d-46e1-a640-7e020318e456",
        "version": 1,
        "content_hash": "sha256:c748075d185006cd08fcb7a91757d1661052f01f0bcb366c092b10c261594660"
      },
      "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": "Paul Barham",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Aakanksha Chowdhery",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Jeff Dean",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Sanjay Ghemawat",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Steven Hand",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Dan Hurt",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Michael Isard",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Hyeontaek Lim",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Ruoming Pang",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Sudip Roy",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Brennan Saeta",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Parker Schuh",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Ryan Sepassi",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Laurent El Shafey",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Chandramohan A. Thekkath",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Yonghui Wu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
            "url": "https://arxiv.org/abs/2203.12533v1"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "The orchestration-system contribution is distinct from PaLM's language-model training. Inherits dataflow/SPMD/MPMD systems; measured utilization on selected TPU workloads is not universal100percent utilization, failure freedom or proof of optimal resource scheduling.",
        "domain": "Large-scale AI orchestration systems"
      },
      "sources": [
        {
          "citation": "Pathways: Asynchronous Distributed Dataflow for ML",
          "url": "https://arxiv.org/abs/2203.12533v1"
        }
      ],
      "statement": "# Pathways: asynchronous distributed accelerator dataflow\n\nUsed sharded asynchronous operators over futures and gang scheduling to decouple control-plane execution from dependent accelerator work under a single-controller programming model.\n\n## Scientific scope\n\nThe orchestration-system contribution is distinct from PaLM's language-model training. Inherits dataflow/SPMD/MPMD systems; measured utilization on selected TPU workloads is not universal100percent utilization, failure freedom or proof of optimal resource scheduling.\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\nPaul Barham; Aakanksha Chowdhery; Jeff Dean; Sanjay Ghemawat; Steven Hand; Dan Hurt; Michael Isard; Hyeontaek Lim; Ruoming Pang; Sudip Roy; Brennan Saeta; Parker Schuh; Ryan Sepassi; Laurent El Shafey; Chandramohan A. Thekkath; Yonghui Wu\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- [Pathways: Asynchronous Distributed Dataflow for ML](https://arxiv.org/abs/2203.12533v1)\n",
      "summary": "Used sharded asynchronous operators over futures and gang scheduling to decouple control-plane execution from dependent accelerator work under a single-controller programming model.",
      "title": "Pathways: asynchronous distributed accelerator dataflow"
    },
    "record_id": "urn:uuid:b84cb4f0-821d-46e1-a640-7e020318e456",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:b84cb4f0-821d-46e1-a640-7e020318e456",
      "version": 1,
      "content_hash": "sha256:c748075d185006cd08fcb7a91757d1661052f01f0bcb366c092b10c261594660"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:958b49c571177e4053f498e49c2fec63f9a4b07c77fa4f0a1e686eff97bb2ac2"
  },
  "target": {
    "record_id": "urn:uuid:b84cb4f0-821d-46e1-a640-7e020318e456",
    "version": 1,
    "content_hash": "sha256:c748075d185006cd08fcb7a91757d1661052f01f0bcb366c092b10c261594660"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "barham-et-al-pathways-asynchronous-dataflow-2022",
    "entry_id": "barham-et-al-pathways-asynchronous-dataflow-2022",
    "revision": 1,
    "title": "Pathways: asynchronous distributed accelerator dataflow",
    "context": "Large-scale AI orchestration systems",
    "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": "Used sharded asynchronous operators over futures and gang scheduling to decouple control-plane execution from dependent accelerator work under a single-controller programming model.",
    "scope": "Used sharded asynchronous operators over futures and gang scheduling to decouple control-plane execution from dependent accelerator work under a single-controller programming model. The orchestration-system contribution is distinct from PaLM's language-model training. Inherits dataflow/SPMD/MPMD systems; measured utilization on selected TPU workloads is not universal100percent utilization, failure freedom or proof of optimal resource scheduling.",
    "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. The orchestration-system contribution is distinct from PaLM's language-model training. Inherits dataflow/SPMD/MPMD systems; measured utilization on selected TPU workloads is not universal100percent utilization, failure freedom or proof of optimal resource scheduling.",
    "references": [
      {
        "title": "Pathways: Asynchronous Distributed Dataflow for ML",
        "url": "https://arxiv.org/abs/2203.12533v1",
        "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:b84cb4f0-821d-46e1-a640-7e020318e456",
      "version": 1,
      "content_hash": "sha256:c748075d185006cd08fcb7a91757d1661052f01f0bcb366c092b10c261594660"
    },
    "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"
}