{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "bytedance-seed-streaming-mixedreward-training-2025",
    "title": "ByteDance Seed: streaming rollout and mixed-reward reasoning training",
    "summary": "Integrated reasoning-based answer verification and pairwise preference rewards with online prompt-distribution adaptation and a streaming rollout scheduler that continues partial generations across policy versions.",
    "boundary": "VAPO, DAPO, HybridFlow, KARP and ByteCheckpoint are inherited or separately described methods, not newly invented by this integration. LLM verification is fallible; partial off-policy generations introduce policy-lag concerns. Internal benchmark results are source-reported, and promised benchmark release is not verified here.",
    "area": "Reasoning-model reinforcement-learning systems",
    "topic": "Artificial intelligence",
    "year": 2025,
    "lane": "HUMAN",
    "result": "EMPIRICAL",
    "package_id": "bytedance-seed-streaming-mixedreward-training-2025",
    "target": {
      "record_id": "urn:uuid:82b0c240-b684-4831-a161-19197fffa449",
      "version": 1,
      "content_hash": "sha256:5ddf37814a003f4a7833cbef570d104058cd7e191f4afa52ca64446dffa07376"
    },
    "humans": [],
    "authors_complete": false,
    "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-2504-13914v1",
        "title": "Seed-Thinking-v1.5: Advancing Superb Reasoning Models with Reinforcement Learning",
        "url": "https://arxiv.org/abs/2504.13914v1"
      },
      {
        "id": "cur20261015-support-bytedance-seed-streaming-mixedreward-training-2025",
        "title": "ByteDance Seed: streaming rollout and mixed-reward reasoning training — official research provenance",
        "url": "https://seed.bytedance.com/en/blog/bytedance-s-latest-thinking-model-seed-thinking-v1-5-technical-details-disclosed"
      }
    ],
    "points": "75000",
    "published_grade": {
      "id": "bytedance-seed-streaming-mixedreward-training-2025",
      "entry_id": "bytedance-seed-streaming-mixedreward-training-2025",
      "band": "B",
      "modifier": "+",
      "target": {
        "record_id": "urn:uuid:82b0c240-b684-4831-a161-19197fffa449",
        "version": 1,
        "content_hash": "sha256:5ddf37814a003f4a7833cbef570d104058cd7e191f4afa52ca64446dffa07376"
      },
      "what_changed": "Integrated reasoning-based answer verification and pairwise preference rewards with online prompt-distribution adaptation and a streaming rollout scheduler that continues partial generations across policy versions.",
      "scope": "Integrated reasoning-based answer verification and pairwise preference rewards with online prompt-distribution adaptation and a streaming rollout scheduler that continues partial generations across policy versions. VAPO, DAPO, HybridFlow, KARP and ByteCheckpoint are inherited or separately described methods, not newly invented by this integration. LLM verification is fallible; partial off-policy generations introduce policy-lag concerns. Internal benchmark results are source-reported, and promised benchmark release is not verified here.",
      "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:82b0c240-b684-4831-a161-19197fffa449",
        "version": 1,
        "content_hash": "sha256:5ddf37814a003f4a7833cbef570d104058cd7e191f4afa52ca64446dffa07376"
      },
      "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": "organization",
          "name": "ByteDance Seed",
          "note": "Cited historical collaboration, not a platform organization account; individual attribution remains incomplete.",
          "roles": [
            "discoverer"
          ],
          "source": {
            "citation": "Seed-Thinking-v1.5: Advancing Superb Reasoning Models with Reinforcement Learning",
            "url": "https://arxiv.org/abs/2504.13914v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "Seed-Thinking-v1.5: Advancing Superb Reasoning Models with Reinforcement Learning",
            "url": "https://arxiv.org/abs/2504.13914v1"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "ByteDance Seed: streaming rollout and mixed-reward reasoning training — official research provenance",
            "url": "https://seed.bytedance.com/en/blog/bytedance-s-latest-thinking-model-seed-thinking-v1-5-technical-details-disclosed"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "VAPO, DAPO, HybridFlow, KARP and ByteCheckpoint are inherited or separately described methods, not newly invented by this integration. LLM verification is fallible; partial off-policy generations introduce policy-lag concerns. Internal benchmark results are source-reported, and promised benchmark release is not verified here.",
        "domain": "Reasoning-model reinforcement-learning systems"
      },
      "sources": [
        {
          "citation": "Seed-Thinking-v1.5: Advancing Superb Reasoning Models with Reinforcement Learning",
          "url": "https://arxiv.org/abs/2504.13914v1"
        },
        {
          "citation": "ByteDance Seed: streaming rollout and mixed-reward reasoning training — official research provenance",
          "url": "https://seed.bytedance.com/en/blog/bytedance-s-latest-thinking-model-seed-thinking-v1-5-technical-details-disclosed"
        }
      ],
      "statement": "# ByteDance Seed: streaming rollout and mixed-reward reasoning training\n\nIntegrated reasoning-based answer verification and pairwise preference rewards with online prompt-distribution adaptation and a streaming rollout scheduler that continues partial generations across policy versions.\n\n## Scientific scope\n\nVAPO, DAPO, HybridFlow, KARP and ByteCheckpoint are inherited or separately described methods, not newly invented by this integration. LLM verification is fallible; partial off-policy generations introduce policy-lag concerns. Internal benchmark results are source-reported, and promised benchmark release is not verified here.\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\n\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- [Seed-Thinking-v1.5: Advancing Superb Reasoning Models with Reinforcement Learning](https://arxiv.org/abs/2504.13914v1)\n- [ByteDance Seed: streaming rollout and mixed-reward reasoning training — official research provenance](https://seed.bytedance.com/en/blog/bytedance-s-latest-thinking-model-seed-thinking-v1-5-technical-details-disclosed)\n",
      "summary": "Integrated reasoning-based answer verification and pairwise preference rewards with online prompt-distribution adaptation and a streaming rollout scheduler that continues partial generations across policy versions.",
      "title": "ByteDance Seed: streaming rollout and mixed-reward reasoning training"
    },
    "record_id": "urn:uuid:82b0c240-b684-4831-a161-19197fffa449",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:82b0c240-b684-4831-a161-19197fffa449",
      "version": 1,
      "content_hash": "sha256:5ddf37814a003f4a7833cbef570d104058cd7e191f4afa52ca64446dffa07376"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:9863f14acecb6e2c8ddc17774cea347457a1a67feb143debdfa7c5d8aaa18ec4"
  },
  "target": {
    "record_id": "urn:uuid:82b0c240-b684-4831-a161-19197fffa449",
    "version": 1,
    "content_hash": "sha256:5ddf37814a003f4a7833cbef570d104058cd7e191f4afa52ca64446dffa07376"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "bytedance-seed-streaming-mixedreward-training-2025",
    "entry_id": "bytedance-seed-streaming-mixedreward-training-2025",
    "revision": 1,
    "title": "ByteDance Seed: streaming rollout and mixed-reward reasoning training",
    "context": "Reasoning-model reinforcement-learning systems",
    "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": "Integrated reasoning-based answer verification and pairwise preference rewards with online prompt-distribution adaptation and a streaming rollout scheduler that continues partial generations across policy versions.",
    "scope": "Integrated reasoning-based answer verification and pairwise preference rewards with online prompt-distribution adaptation and a streaming rollout scheduler that continues partial generations across policy versions. VAPO, DAPO, HybridFlow, KARP and ByteCheckpoint are inherited or separately described methods, not newly invented by this integration. LLM verification is fallible; partial off-policy generations introduce policy-lag concerns. Internal benchmark results are source-reported, and promised benchmark release is not verified here.",
    "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. VAPO, DAPO, HybridFlow, KARP and ByteCheckpoint are inherited or separately described methods, not newly invented by this integration. LLM verification is fallible; partial off-policy generations introduce policy-lag concerns. Internal benchmark results are source-reported, and promised benchmark release is not verified here.",
    "references": [
      {
        "title": "Seed-Thinking-v1.5: Advancing Superb Reasoning Models with Reinforcement Learning",
        "url": "https://arxiv.org/abs/2504.13914v1",
        "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification."
      },
      {
        "title": "ByteDance Seed: streaming rollout and mixed-reward reasoning training — official research provenance",
        "url": "https://seed.bytedance.com/en/blog/bytedance-s-latest-thinking-model-seed-thinking-v1-5-technical-details-disclosed",
        "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:82b0c240-b684-4831-a161-19197fffa449",
      "version": 1,
      "content_hash": "sha256:5ddf37814a003f4a7833cbef570d104058cd7e191f4afa52ca64446dffa07376"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "75000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}