{
  "schema_version": "2.0",
  "record_type": "dataset",
  "tier": "verified",
  "slug": "disc-med-sft",
  "name": "DISC-Med-SFT",
  "summary": {
    "zh": "许可证声明：有限制（NOASSERTION）；本站处理：只链接原出处。许可证有限制：自身标注 Apache-2.0 比上游更宽，按最严格的上游（cMedQA2 限定非商业研究；MedDialog 版权归 haodf.com）判受限；只链接原出处。",
    "en": "License: restricted (NOASSERTION); Mdexa hosting: link-only. Declared Apache-2.0 is broader than its upstream; the strictest upstream applies (cMedQA2 non-commercial research only; MedDialog copyright belongs to haodf.com); link only."
  },
  "systems": [
    "modern-medicine"
  ],
  "domains": [
    "clinical-medicine"
  ],
  "content_types": [
    "qa-dialogue"
  ],
  "content_description": {
    "en": "Consultation and knowledge QA fine-tuning data",
    "zh": "问诊与知识问答微调数据"
  },
  "languages": [
    "zh"
  ],
  "modalities": [
    "text"
  ],
  "size": {
    "raw": "GitHub 称近47万条；HF 子集 464,898行",
    "measures": [
      {
        "value": 470000,
        "unit": "items",
        "qualifier": "approx",
        "scope": "GitHub"
      },
      {
        "value": 464898,
        "unit": "rows",
        "qualifier": "exact",
        "scope": "HF subset"
      }
    ]
  },
  "publisher": "复旦大学 DISC 实验室（FudanDISC）",
  "year": {
    "value": 2023,
    "raw": "2023（arXiv 2308.14346）"
  },
  "expert_review": {
    "en": "2k human-curated preference items; not stated for the rest",
    "zh": "人工整理偏好数据2k条；其余未说明"
  },
  "license": {
    "legacy_status": "pending",
    "spdx": "NOASSERTION",
    "commercial_use": "no",
    "derivatives": "unstated",
    "redistribution": "unstated",
    "rationale": {
      "zh": "自身标注 Apache-2.0，但上游 cMedQA2 限定非商业研究（\"available for non-commercial research\"），上游 MedDialog 的镜像页称数据版权归 haodf.com；按\"取所有上游里最严格的\"判受限",
      "en": "Declares Apache-2.0, but upstream cMedQA2 is limited to non-commercial research and a MedDialog mirror says copyright belongs to haodf.com; the strictest upstream applies: restricted"
    },
    "evidence": [
      {
        "location": "hf-card-metadata",
        "location_detail": {
          "en": "HF dataset card: Apache-2.0",
          "zh": "HF 数据集卡：Apache-2.0"
        },
        "excerpt": "license: apache-2.0",
        "spdx": "Apache-2.0",
        "url": "https://huggingface.co/datasets/Flmc/DISC-Med-SFT",
        "verified_on": "2026-11-12",
        "snapshot_id": "s01",
        "observed": "page",
        "secondhand": false
      },
      {
        "location": "repo-license",
        "location_detail": {
          "en": "GitHub repository Apache-2.0; no separate data license",
          "zh": "GitHub 仓库 Apache-2.0；未单独声明数据许可"
        },
        "excerpt": "Apache License\nVersion 2.0, January 2004",
        "spdx": "Apache-2.0",
        "url": "https://github.com/FudanDISC/DISC-MedLLM",
        "verified_on": "2026-11-12",
        "snapshot_id": "s04",
        "observed": "page",
        "secondhand": false
      },
      {
        "location": "upstream-terms",
        "location_detail": {
          "zh": "上游 cMedQA2 的 README（见 cmedqa2 卡片快照 s01）：限定非商业研究",
          "en": "Upstream cMedQA2 README (cmedqa2 card, snapshot s01): non-commercial research only"
        },
        "excerpt": "available for non-commercial research",
        "spdx": null,
        "url": "https://github.com/zhangsheng93/cMedQA2",
        "verified_on": "2026-11-12",
        "snapshot_id": null,
        "observed": "page",
        "secondhand": false
      },
      {
        "location": "upstream-terms",
        "location_detail": {
          "zh": "上游 MedDialog 的第三方镜像页（见 meddialog-cn 卡片），二手",
          "en": "Upstream MedDialog third-party mirror (meddialog-cn card); secondhand"
        },
        "excerpt": "All copyrights of the data belong to haodf.com.",
        "spdx": null,
        "url": "https://huggingface.co/datasets/UCSD26/medical_dialog",
        "verified_on": "2026-11-12",
        "snapshot_id": null,
        "observed": "secondary-source",
        "secondhand": true
      }
    ],
    "declared_broader_than_upstream": true
  },
  "license_status": "restricted",
  "hosting": "link-only",
  "hosting_reason": {
    "zh": "许可证有限制：自身标注比上游更宽，按最严格的上游（cMedQA2 限定非商业研究；MedDialog 版权归 haodf.com）判受限；只链接原出处。",
    "en": "License is restricted: the declared license is broader than upstream; the strictest upstream applies (cMedQA2 non-commercial; MedDialog copyright haodf.com); link only."
  },
  "signoff": null,
  "repo_license": {
    "spdx": "Apache-2.0",
    "source": "license-file",
    "verified": true,
    "snapshot_id": "s04",
    "url": "https://github.com/FudanDISC/DISC-MedLLM"
  },
  "risk_flags": [
    "derived",
    "scraped",
    "real-patient-text",
    "model-generated"
  ],
  "provenance": {
    "source_description": {
      "en": "Restructured from existing datasets: MedDialog 400k, cMedQA2 20k, CMeKG 50k, human preference data 2k, MedMCQA 8k, MOSS-SFT 33k, Alpaca-GPT4-zh 1k",
      "zh": "由现有数据集重构：MedDialog 400k、cMedQA2 20k、CMeKG 50k、人工偏好数据2k、MedMCQA 8k、MOSS-SFT 33k、Alpaca-GPT4-zh 1k"
    },
    "upstream": [
      {
        "slug": "meddialog-cn",
        "name": "MedDialog",
        "relation": "includes"
      },
      {
        "slug": "cmedqa2",
        "name": "cMedQA2",
        "relation": "includes"
      },
      {
        "slug": null,
        "name": "CMeKG",
        "relation": "includes"
      },
      {
        "slug": null,
        "name": "MedMCQA",
        "relation": "includes"
      },
      {
        "slug": null,
        "name": "MOSS-SFT",
        "relation": "includes"
      },
      {
        "slug": null,
        "name": "Alpaca-GPT4-zh",
        "relation": "includes"
      }
    ]
  },
  "known_issues": [
    {
      "en": "Upstream includes MedDialog (copyright belongs to haodf.com) and cMedQA2 (non-commercial research statement), inconsistent with the Apache-2.0 label",
      "zh": "上游包含 MedDialog（版权归 haodf.com）与 cMedQA2（非商业研究声明），与 Apache-2.0 标注不一致"
    },
    {
      "en": "HF has only a subset (without the behavior preference data)",
      "zh": "HF 仅为子集（不含行为偏好数据）"
    },
    {
      "zh": "部分数据由大模型重建或构造（快照 s03 原文：「* **多轮对话的问询能力**，我们以真实咨询对话纪录作为信息源，使用大模型进行对话重建，构建过程中要求模型完全对齐对话中的医学信息。」「* **可靠丰富的专业知识**，我们以医学知识图谱作为信息源，通过采样三元组，并使用通用大模型的语言能力进行对话样本的构造。」）",
      "en": "Part of the data was reconstructed or generated by an LLM (snapshot s03, quoted in the zh text)."
    }
  ],
  "preflight": {
    "status": "not-run"
  },
  "access": {
    "mdexa_download": null,
    "external_urls": [
      "https://huggingface.co/datasets/Flmc/DISC-Med-SFT",
      "https://github.com/FudanDISC/DISC-MedLLM"
    ],
    "notes": null
  },
  "citation": {
    "text": null,
    "bibtex": null,
    "references": [
      "arXiv:2308.14346"
    ]
  },
  "verification_level": "verified",
  "last_verified": "2026-11-12",
  "tags": [
    {
      "facet": "intended_use",
      "value": "fine-tuning",
      "basis": "author-stated",
      "excerpt": "我们总共发布了近47万条训练数据，其中包括重构AI医患对话和知识图谱问答对。您可以访问这个[链接](https://huggingface.co/datasets/Flmc/DISC-Med-SFT)下载数据集。",
      "snapshot_id": "s03"
    },
    {
      "facet": "task_format",
      "value": "multi-turn-dialogue",
      "basis": "author-stated",
      "excerpt": "* **多轮对话的问询能力**，我们以真实咨询对话纪录作为信息源，使用大模型进行对话重建，构建过程中要求模型完全对齐对话中的医学信息。",
      "snapshot_id": "s03"
    },
    {
      "facet": "task_format",
      "value": "open-qa",
      "basis": "inferred",
      "inferred_from": "card-content-description",
      "note": {
        "en": "Includes single-turn QA subsets.",
        "zh": "含单轮问答子集"
      }
    },
    {
      "facet": "specialty",
      "value": "multi-specialty",
      "basis": "inferred",
      "inferred_from": "card-content-description",
      "note": {
        "zh": "综合问诊与知识问答，未限定专科",
        "en": "综合问诊与知识问答; not limited to a specialty"
      }
    }
  ],
  "changelog": [
    {
      "date": "2026-11-12",
      "change": {
        "en": "Card converted from an earlier manual catalog; no new facts added; reclassified under taxonomy v2.",
        "zh": "由早期人工整理的目录转换生成，未新增事实；按分类体系 v2 重新归类。"
      }
    },
    {
      "date": "2026-11-12",
      "change": {
        "en": "License status split into license_status (what the license says) and hosting (what Mdexa does); the earlier conclusion is kept for traceability.",
        "zh": "把许可证状态拆成 license_status（许可证怎么写）与 hosting（本站怎么处理）；原结论保留以便追溯。"
      }
    },
    {
      "date": "2026-11-12",
      "change": {
        "en": "Added verbatim license excerpts and snapshot ids from captured page snapshots; no conclusion changed.",
        "zh": "补入来自已抓取页面快照的许可证逐字摘录与快照编号；未改动任何结论。"
      }
    },
    {
      "date": "2026-11-12",
      "change": {
        "en": "Recorded the code repository license separately; it is not treated as a data license statement.",
        "zh": "单独记录代码仓库许可证，不视为数据许可声明。"
      }
    },
    {
      "date": "2026-11-13",
      "change": {
        "en": "Added dataset-level tags: 4 (2 stated by the authors with excerpts, 2 inferred from the content). No conclusion changed.",
        "zh": "新增数据集标签 4 个（作者声明 2 个，附逐字摘录；据内容推断 2 个）；未改动任何结论。"
      }
    },
    {
      "date": "2026-11-13",
      "change": {
        "en": "Removed the stored size-tier tag; the tier is now computed from the size field when pages are built.",
        "zh": "移除存储的规模档位标签；档位改为生成页面时由 size 字段计算。"
      }
    },
    {
      "date": "2026-11-13",
      "change": {
        "en": "Added the model-generated risk flag (the dialogues were rebuilt with a language model). Status and hosting unchanged.",
        "zh": "新增 model-generated 风险标记（对话由大模型改写重构）；状态与处理方式不变。"
      }
    },
    {
      "date": "2026-11-13",
      "change": {
        "en": "Added specialty tags: multi-specialty(inferred).",
        "zh": "新增专科标签：multi-specialty(inferred)。"
      }
    },
    {
      "date": "2026-11-13",
      "change": {
        "en": "License status changed from conflict to restricted and hosting from pending to link-only: the declared license is broader than its upstream sources, so the strictest upstream applies.",
        "zh": "许可状态由冲突改为受限，处理方式由待确认改为只链接原出处：数据集标注的许可比上游更宽，按最严格的上游判定。"
      }
    }
  ],
  "access_recipe": {
    "verified_on": "2026-11-14",
    "platform": "huggingface",
    "repo_id": "Flmc/DISC-Med-SFT",
    "source": {
      "api_url": "https://huggingface.co/api/datasets/Flmc/DISC-Med-SFT",
      "fetched_on": "2026-11-14"
    },
    "access_state": "open",
    "files": [
      {
        "name": ".gitattributes",
        "url": "https://huggingface.co/datasets/Flmc/DISC-Med-SFT/resolve/b10eedfc23530ebe257a149b00db73c4162fbfc3/.gitattributes",
        "size_bytes": 2371,
        "checksum": null,
        "format": "gitattributes"
      },
      {
        "name": "DISC-Med-SFT_released.jsonl",
        "url": "https://huggingface.co/datasets/Flmc/DISC-Med-SFT/resolve/b10eedfc23530ebe257a149b00db73c4162fbfc3/DISC-Med-SFT_released.jsonl",
        "size_bytes": 820422846,
        "checksum": {
          "algorithm": "sha256",
          "value": "34beda9ccf9a7a59399d853c47b378b8212b41df6f80f67b58502074daa9d545"
        },
        "format": "jsonl"
      },
      {
        "name": "README.md",
        "url": "https://huggingface.co/datasets/Flmc/DISC-Med-SFT/resolve/b10eedfc23530ebe257a149b00db73c4162fbfc3/README.md",
        "size_bytes": 298,
        "checksum": null,
        "format": "md"
      }
    ],
    "files_truncated": false,
    "commands": [
      {
        "command": "hf download Flmc/DISC-Med-SFT --repo-type dataset",
        "evidence": {
          "excerpt": "You can use the hf download command from the terminal to directly download files from the Hub.",
          "doc": "hf-download",
          "url": "https://huggingface.co/docs/huggingface_hub/guides/download",
          "fetched_on": "2026-11-14"
        }
      }
    ],
    "registration": null,
    "version": {
      "commit": "b10eedfc23530ebe257a149b00db73c4162fbfc3",
      "last_modified": "2023-08-29T12:54:14.000Z"
    },
    "license_note": {
      "license_status": "restricted",
      "text": "Use according to the terms of the official source; this recipe does not say the data can be used freely."
    },
    "gaps": []
  },
  "_links": {
    "text": "在本库继续查找。本条目是 109 条已处理并核实的记录之一（数据集、数据资源与开源模型）。查找其他条目：调用 `https://mcp.mdexa.com/v1/datasets?q=<关键词>`，或连接 MCP 服务 `https://mcp.mdexa.com/mcp`；每分钟 20 次以内无需密钥。每行返回不到 1 KB；空结果即\"数据库未收录\"。核对一份清单：一次批量核对（免费账号）即返回报告。说明：/agents.md。相似条目：CMtMedQA (https://mdexa.com/zh/datasets/cmtmedqa/)；Huatuo-Llama-Med-Chinese 知识库问答与文献数据 (https://mdexa.com/zh/datasets/huatuo-llama-med-chinese-data/)；BianQueCorpus（扁鹊健康大数据） (https://mdexa.com/zh/datasets/bianque-corpus/)。",
    "records": 109,
    "search": {
      "template": "https://mcp.mdexa.com/v1/datasets?q={words}",
      "example": "https://mcp.mdexa.com/v1/datasets?q=%E7%BB%93%E5%90%88%E4%BA%B2%E5%92%8C%E5%8A%9B"
    },
    "mcp": "https://mcp.mdexa.com/mcp",
    "anonymous_calls_per_minute": 20,
    "guide": "https://mdexa.com/zh/agents.md",
    "faq": "https://mdexa.com/zh/developers/faq/",
    "similar": [
      {
        "name": "CMtMedQA",
        "url": "https://mdexa.com/zh/datasets/cmtmedqa/",
        "json": "https://mdexa.com/zh/datasets/cmtmedqa.json"
      },
      {
        "name": "Huatuo-Llama-Med-Chinese 知识库问答与文献数据",
        "url": "https://mdexa.com/zh/datasets/huatuo-llama-med-chinese-data/",
        "json": "https://mdexa.com/zh/datasets/huatuo-llama-med-chinese-data.json"
      },
      {
        "name": "BianQueCorpus（扁鹊健康大数据）",
        "url": "https://mdexa.com/zh/datasets/bianque-corpus/",
        "json": "https://mdexa.com/zh/datasets/bianque-corpus.json"
      }
    ]
  }
}