{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "provilkov-emelianenko-voita-bpe-dropout-2019",
    "title": "BPE-dropout: stochastic merge suppression for subword training",
    "summary": "Kept a learned BPE merge table while randomly suppressing candidate merges during training segmentation, exposing the translation model to multiple segmentations before deterministic inference.",
    "boundary": "Does not introduce a new BPE vocabulary or establish universal robustness. Dropout acts on merge operations, not neural activations. It inherits subword regularization's motivation and uses standard BPE when the dropout probability is zero.",
    "area": "Tokenization and regularization",
    "topic": "Artificial intelligence",
    "year": 2019,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "provilkov-emelianenko-voita-bpe-dropout-2019",
    "target": {
      "record_id": "urn:uuid:f551e298-7fae-46aa-ac99-90a06a2478c4",
      "version": 1,
      "content_hash": "sha256:fee2b8911d4fb72e10f3dca5538b0b6e6c09fad76c0a72150193ae60fbd603b8"
    },
    "humans": [
      "Ivan Provilkov",
      "Dmitrii Emelianenko",
      "Elena Voita"
    ],
    "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-1910-13267v1",
        "title": "BPE-Dropout: Simple and Effective Subword Regularization",
        "url": "https://arxiv.org/abs/1910.13267v1"
      }
    ],
    "points": "70000",
    "published_grade": {
      "id": "provilkov-emelianenko-voita-bpe-dropout-2019",
      "entry_id": "provilkov-emelianenko-voita-bpe-dropout-2019",
      "band": "B",
      "modifier": "+",
      "target": {
        "record_id": "urn:uuid:f551e298-7fae-46aa-ac99-90a06a2478c4",
        "version": 1,
        "content_hash": "sha256:fee2b8911d4fb72e10f3dca5538b0b6e6c09fad76c0a72150193ae60fbd603b8"
      },
      "what_changed": "Kept a learned BPE merge table while randomly suppressing candidate merges during training segmentation, exposing the translation model to multiple segmentations before deterministic inference.",
      "scope": "Kept a learned BPE merge table while randomly suppressing candidate merges during training segmentation, exposing the translation model to multiple segmentations before deterministic inference. Does not introduce a new BPE vocabulary or establish universal robustness. Dropout acts on merge operations, not neural activations. It inherits subword regularization's motivation and uses standard BPE when the dropout probability is zero.",
      "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:f551e298-7fae-46aa-ac99-90a06a2478c4",
        "version": 1,
        "content_hash": "sha256:fee2b8911d4fb72e10f3dca5538b0b6e6c09fad76c0a72150193ae60fbd603b8"
      },
      "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": "Ivan Provilkov",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "BPE-Dropout: Simple and Effective Subword Regularization",
            "url": "https://arxiv.org/abs/1910.13267v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Dmitrii Emelianenko",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "BPE-Dropout: Simple and Effective Subword Regularization",
            "url": "https://arxiv.org/abs/1910.13267v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Elena Voita",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "BPE-Dropout: Simple and Effective Subword Regularization",
            "url": "https://arxiv.org/abs/1910.13267v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "BPE-Dropout: Simple and Effective Subword Regularization",
            "url": "https://arxiv.org/abs/1910.13267v1"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "Does not introduce a new BPE vocabulary or establish universal robustness. Dropout acts on merge operations, not neural activations. It inherits subword regularization's motivation and uses standard BPE when the dropout probability is zero.",
        "domain": "Tokenization and regularization"
      },
      "sources": [
        {
          "citation": "BPE-Dropout: Simple and Effective Subword Regularization",
          "url": "https://arxiv.org/abs/1910.13267v1"
        }
      ],
      "statement": "# BPE-dropout: stochastic merge suppression for subword training\n\nKept a learned BPE merge table while randomly suppressing candidate merges during training segmentation, exposing the translation model to multiple segmentations before deterministic inference.\n\n## Scientific scope\n\nDoes not introduce a new BPE vocabulary or establish universal robustness. Dropout acts on merge operations, not neural activations. It inherits subword regularization's motivation and uses standard BPE when the dropout probability is zero.\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\nIvan Provilkov; Dmitrii Emelianenko; Elena Voita\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- [BPE-Dropout: Simple and Effective Subword Regularization](https://arxiv.org/abs/1910.13267v1)\n",
      "summary": "Kept a learned BPE merge table while randomly suppressing candidate merges during training segmentation, exposing the translation model to multiple segmentations before deterministic inference.",
      "title": "BPE-dropout: stochastic merge suppression for subword training"
    },
    "record_id": "urn:uuid:f551e298-7fae-46aa-ac99-90a06a2478c4",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:f551e298-7fae-46aa-ac99-90a06a2478c4",
      "version": 1,
      "content_hash": "sha256:fee2b8911d4fb72e10f3dca5538b0b6e6c09fad76c0a72150193ae60fbd603b8"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:61eb7e2716cdda1015453b71695d172ab2e5b6a260c5784c9102e1736b5e27a6"
  },
  "target": {
    "record_id": "urn:uuid:f551e298-7fae-46aa-ac99-90a06a2478c4",
    "version": 1,
    "content_hash": "sha256:fee2b8911d4fb72e10f3dca5538b0b6e6c09fad76c0a72150193ae60fbd603b8"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "provilkov-emelianenko-voita-bpe-dropout-2019",
    "entry_id": "provilkov-emelianenko-voita-bpe-dropout-2019",
    "revision": 1,
    "title": "BPE-dropout: stochastic merge suppression for subword training",
    "context": "Tokenization and regularization",
    "track": "Human-designed research method or resource",
    "band": "B",
    "modifier": "+",
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-10-16",
    "evidence_cutoff": "2026-10-16",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Kept a learned BPE merge table while randomly suppressing candidate merges during training segmentation, exposing the translation model to multiple segmentations before deterministic inference.",
    "scope": "Kept a learned BPE merge table while randomly suppressing candidate merges during training segmentation, exposing the translation model to multiple segmentations before deterministic inference. Does not introduce a new BPE vocabulary or establish universal robustness. Dropout acts on merge operations, not neural activations. It inherits subword regularization's motivation and uses standard BPE when the dropout probability is zero.",
    "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. Does not introduce a new BPE vocabulary or establish universal robustness. Dropout acts on merge operations, not neural activations. It inherits subword regularization's motivation and uses standard BPE when the dropout probability is zero.",
    "references": [
      {
        "title": "BPE-Dropout: Simple and Effective Subword Regularization",
        "url": "https://arxiv.org/abs/1910.13267v1",
        "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:f551e298-7fae-46aa-ac99-90a06a2478c4",
      "version": 1,
      "content_hash": "sha256:fee2b8911d4fb72e10f3dca5538b0b6e6c09fad76c0a72150193ae60fbd603b8"
    },
    "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"
}