Refresh CIMD dataset snapshot

.gitattributes

@@ -0,0 +1,56 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+data/text/iron_ore/**/train.jsonl filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,2 @@
+.DS_Store
+._*

OpenCSG数据集社区许可协议.md

@@ -0,0 +1,541 @@
+# OpenCSG 数据集社区许可协议
+
+版本 1.0
+OpenCSG
+
+## 1. 定义
+
+### 1.1 OpenCSG
+
+“OpenCSG”指 OpenCSG Inc. 及其关联公司。
+
+### 1.2 数据集
+
+“数据集”指由 OpenCSG 发布的一个或多个数据集、数据文件、样本、元数据、标签、索引、解析文本、说明文档、版本更新及其相关材料。
+
+### 1.3 原始来源材料
+
+“原始来源材料”指构成或用于构建数据集的公开资料、采购资料、合作方资料、第三方授权资料及其他来源材料。
+
+### 1.4 元数据
+
+“元数据”指与数据集相关的标题、作者、来源、时间、许可标记、路径、标签、分类、标识及其他描述性信息。
+
+### 1.5 增强数据
+
+“增强数据”指被许可方基于数据集进行清洗、筛选、纠错、对齐、标注、翻译、重组、扩充或其他加工后形成的数据，但不包括结果与模型。
+
+### 1.6 结果
+
+“结果”指通过对数据集进行分析、统计、评测、检索、嵌入、聚类、特征提取、训练或其他计算使用而得到的输出、指标、报告、图表、参数、洞察、推断或其他成果。
+
+### 1.7 模型
+
+“模型”指利用数据集进行预训练、继续预训练、微调、对齐、蒸馏、检索增强、评测或其他机器学习活动后得到的模型权重、适配器、提示模板、评测工件、系统配置及其衍生物。
+
+### 1.8 商业用途
+
+“商业用途”指以直接或间接获取商业利益、营业收入、服务收入、客户合同收入、订阅收入、广告收入、佣金、授权费或其他经济利益为目的的使用行为。
+
+### 1.9 再分发
+
+“再分发”指向第三方提供、共享、转让、出售、出租、出借、镜像、托管、公开发布、以 API 或下载方式开放，或以其他方式使第三方获得数据集或其可替代部分的行为。
+
+### 1.10 实质性部分
+
+“实质性部分”指就数量、质量、可替代性、重建可能性或商业价值而言，足以替代数据集整体或重要组成部分的内容。
+
+## 2. 授权授予
+
+### 2.1 有限许可
+
+在被许可方持续遵守本协议的前提下，OpenCSG 向被许可方授予一项非独占、不可转让、不可再许可、可撤销的有限许可。
+
+### 2.2 许可范围
+
+该许可仅允许被许可方：
+
+1. 访问、下载和内部存储数据集。
+2. 为研究、评测、验证、内部开发、模型训练、模型微调、检索增强、质量分析和合规审查而使用数据集。
+3. 在被许可方自身或受其控制的基础设施内处理数据集。
+4. 基于数据集生成结果和模型，但应遵守本协议关于再分发、商业化、来源约束和合规义务的全部限制。
+
+### 2.3 保留权利
+
+除本协议明确授予的权利外，OpenCSG 不向被许可方授予任何明示或默示的其他权利。
+
+## 3. 来源权利与授权边界
+
+### 3.1 权利范围
+
+数据集可能包含来自公开来源、采购来源、合作来源或其他第三方来源的材料。被许可方确认，OpenCSG 仅在其依法享有或有权授予的范围内授予本协议项下许可。
+
+### 3.2 特别限制优先
+
+如数据集中特定文件、子集、字段、样本或来源材料附带更严格的来源说明、许可说明、使用限制或合规要求，则该等更严格条款优先适用。
+
+### 3.3 不构成独立原始授权
+
+本协议不授予被许可方对原始来源材料本身的独立权利，也不授予任何超出 OpenCSG 实际权利边界的再授权、再许可或再分发权利。
+
+### 3.4 不授予其他第三方权利
+
+被许可方不得以本协议为依据主张其有权获得任何第三方商标权、署名权、人格权、数据库权、隐私权、数据权益、商业秘密权或其他未由 OpenCSG 明确授予的权利。
+
+## 4. 使用限制
+
+### 4.1 未经书面授权不得从事的行为
+
+除非获得 OpenCSG 事先书面授权，被许可方不得：
+
+1. 再分发数据集、增强数据或数据集的实质性部分。
+2. 以镜像站、下载站、数据包、SaaS、PaaS、API、托管检索服务、数据接口或其他对外服务方式提供数据集。
+3. 将数据集嵌入、捆绑或交付于商业产品、商业服务、数据产品、数据服务或对外收费解决方案中。
+4. 出售、转售、转授权、出租、出借或交换数据集。
+5. 发布可使第三方重建、提取或替代数据集整体或重要部分的样本集合、切分结果、缓存、索引或中间件。
+
+### 4.2 权利管理与技术限制
+
+被许可方不得：
+
+1. 删除、遮蔽、篡改或伪造数据集中的来源信息、署名信息、版本信息、许可标记、来源路径、版权声明或其他权利管理信息。
+2. 绕过 OpenCSG 设定的访问控制、许可证控制、速率限制、下载限制、功能门控或安全措施。
+3. 将数据集用于构建与 OpenCSG 数据集形成直接替代关系的数据分发服务、数据聚合服务或数据��易服务。
+4. 明示或暗示 OpenCSG 对被许可方、其产品、其服务或其结果提供背书。
+
+### 4.3 禁止违法使用
+
+被许可方不得将数据集用于任何违法、侵权、歧视性、欺诈性、滥用性或违反适用监管要求的场景。
+
+## 5. 结果与模型
+
+### 5.1 结果与模型的生成
+
+被许可方可以基于数据集生成结果和模型。
+
+### 5.2 不得以结果或模型规避限制
+
+结果与模型本身不当然等同于数据集，但被许可方不得以结果或模型为形式规避本协议对数据集、增强数据或实质性部分的限制。
+
+### 5.3 不得发布可逆向还原数据集的工件
+
+被许可方不得发布、交付或商业化任何足以使第三方恢复、抽取、定位、重建或大规模替代数据集内容的结果、模型、权重、索引、缓存、向量库、检索库或其他工件。
+
+### 5.4 允许的公开成果
+
+被许可方可以发布论文、评测结论、统计指标、基准结果和少量非实质性引用，但应保留合理署名，并不得披露超出必要范围的实质性数据内容。
+
+### 5.5 商业模型与系统的补充授权
+
+如被许可方拟将基于数据集训练、微调、蒸馏或增强得到的模型、系统、Agent、API 服务或商业产品对外提供，OpenCSG 可根据数据来源、使用范围、交付方式和商业模式要求其另行签署商业许可或补充协议。
+
+## 6. 署名与通知义务
+
+### 6.1 合理署名
+
+被许可方在公开引用、论文发表、产品说明、评测报告或对外宣传中使用数据集时，应合理注明：
+
+1. 数据集名称。
+2. OpenCSG。
+3. 数据集版本。
+4. 数据集链接或协议链接。
+5. OpenCSG 要求保留的其他署名或来源说明。
+
+### 6.2 修改说明
+
+如被许可方在允许范围内公开增强数据或非实质性片段，应明确说明其已作修改，并保留原有署名、来源说明和许可说明。
+
+### 6.3 权利声明保留
+
+被许可方不得去除或弱化 OpenCSG 的权利声明、免责声明、责任限制、来源说明和许可说明。
+
+## 7. 合规与审计
+
+### 7.1 合规义务
+
+被许可方应自行确保其对数据集的接收、存储、处理、发布、训练、部署和商业化活动符合适用法律法规及监管要求，包括但不限于知识产权、数据合规、个人信息保护、网络安全、出口管制和行业监管规则。
+
+### 7.2 合规说明
+
+如 OpenCSG 基于合理理由认为被许可方存在违规使用、超范围再分发、未经授权商业化、绕过限制或其他重大违约行为，OpenCSG 有权要求被许可方提供必要的合规说明。
+
+### 7.3 审计权
+
+在合理通知下，OpenCSG 有权对涉及本协议重大义务履行情况进行审计、核验或要求提交合规报告。
+
+## 8. 知识产权
+
+### 8.1 权利归属
+
+数据集及其相关文档、元数据结构、版本组织方式、标识体系、说明文本及 OpenCSG 提供的相关材料中的一切权利、所有权和利益，均归 OpenCSG 或其合法权利人所有。
+
+### 8.2 不发生权利转移
+
+除本协议明确授予的有限使用权外，不发生任何知识产权转让。
+
+### 8.3 商标保护
+
+OpenCSG、CSGHub 及相关名称、标识、商标、服务标志和品牌元素均受法律保护，未经书面许可不得使用。
+
+## 9. 免责声明
+
+### 9.1 现状提供
+
+数据集按“现状”（AS IS）提供。
+
+### 9.2 不作保证
+
+在适用法律允许的最大范围内，OpenCSG 不就数据集作出任何明示或默示保证，包括但不限于权属、完整性、持续可用性、准确性、合法性、适销性、特定用途适用性、无侵权、无错误、无中断或无安全缺陷的保证。
+
+### 9.3 风险自担
+
+被许可方应自行判断数据集是否适合其具体用途，并自行承担相应风险。
+
+## 10. 责任限制
+
+### 10.1 间接损失
+
+在适用法律允许的最大范围内，OpenCSG 不对任何间接损失、附带损失、特殊损失、惩罚性损失或后果性损失承担责任。
+
+### 10.2 业务与数据损失
+
+OpenCSG 不对业务损失、利润损失、商誉损失、机会损失、数据损失、模型损失、中断损失或第三方索赔承担责任。
+
+### 10.3 责任范围
+
+无论基于合同、侵权、严格责任或其他法律理论，OpenCSG 在本协议项下的责任均以适用法律允许的最低范围为限。
+
+## 11. 终止
+
+### 11.1 终止事由
+
+如被许可方违反本协议任何条款，或从事未经授权的再分发、商业化、规避限制、侵权使用或其他重大违约行为，本协议项下许可可由 OpenCSG 立即终止或自动终止。
+
+### 11.2 终止后的义务
+
+终止后，被许可方应立即：
+
+1. 停止接收、使用、处理和对外提供数据集。
+2. 删除或销毁其控制下的数据集及其未获授权保留的副本。
+3. 根据 OpenCSG 要求出具删除或停止使用的书面确认。
+
+### 11.3 可继续保留的内容
+
+但在终止前已合法形成且不包含数据集实质性部分的结果、论文和合规保留记录，可在不违反本协议其他条款的前提下继续保留。
+
+## 12. 商业授权
+
+### 12.1 需另行取得授权的场景
+
+如涉及以下任一场景，被许可方应事先联系 OpenCSG 获取单独的商业许可、OEM 许可、分发许可或补充授权：
+
+1. 再分发数据集。
+2. 向第三方提供数据集下载、镜像、托管、检索、调用或访问服务。
+3. 将数据集嵌入商业产品、商业服务、SaaS、PaaS、API 或付费方案。
+4. 对外提供基于数据集训练、微调、蒸馏、增强或显著受其支撑的模型、Agent、服务或解决方案。
+5. 转售、捆绑、分包、贴牌、OEM、渠道分发或行业解决方案交付。
+
+### 12.2 联系方式
+
+商业授权申请联系方式：
+lorraineg@opencsg.com
+
+## 13. 适用法律与争议解决
+
+### 13.1 适用法律
+
+本协议的订立、效力、解释、履行及争议解决，适用 OpenCSG 指定主体所在地法律；默认可设为中华人民共和国法律。
+
+### 13.2 可选法域
+
+如 OpenCSG 出于集团架构或交易安排需要，也可在正式版本中指定为新加坡、开曼群岛或其他适当法域。
+
+### 13.3 争议解决
+
+因本协议引起或与本协议有关的争议，双方应先友好协商；协商不成的，提交协议正式版本指定的法院或仲裁机构解决。
+
+## 14. 其他
+
+### 14.1 协议更新
+
+OpenCSG 可在合理范围内发布本协议的新版本。除非另有说明，被许可方继续使用数据集即表示接受适用于后续访问或后续版本的新条款。
+
+### 14.2 条款可分割性
+
+如本协议任何条款被认定为无效或不可执行，不影响其他条款的效力。
+
+### 14.3 完整协议
+
+本协议构成被许可方与 OpenCSG 就数据集使用事项达成的完整协议。
+
+---
+
+# OpenCSG Dataset Community License Agreement
+
+Version 1.0
+OpenCSG
+
+## 1. Definitions
+
+### 1.1 OpenCSG
+
+"OpenCSG" means OpenCSG Inc. and its affiliates.
+
+### 1.2 Dataset
+
+"Dataset" means one or more datasets, data files, samples, metadata, labels, indexes, parsed text, documentation, version updates, and related materials published by OpenCSG.
+
+### 1.3 Source Materials
+
+"Source Materials" means public materials, procured materials, partner materials, third-party licensed materials, and other source materials that constitute or are used to build the Dataset.
+
+### 1.4 Metadata
+
+"Metadata" means titles, authors, sources, timestamps, license markings, paths, tags, classifications, identifiers, and other descriptive information associated with the Dataset.
+
+### 1.5 Enhanced Data
+
+"Enhanced Data" means data formed by the Licensee through cleaning, filtering, correction, alignment, annotation, translation, reorganization, enrichment, or other processing based on the Dataset, but excludes Results and Models.
+
+### 1.6 Results
+
+"Results" means outputs, metrics, reports, charts, parameters, insights, inferences, and other deliverables produced through analysis, statistics, evaluation, retrieval, embedding, clustering, feature extraction, training, or other computational use of the Dataset.
+
+### 1.7 Models
+
+"Models" means model weights, adapters, prompt templates, evaluation artifacts, system configurations, and their derivatives obtained through pre-training, continued pre-training, fine-tuning, alignment, distillation, retrieval augmentation, evaluation, or other machine learning activities using the Dataset.
+
+### 1.8 Commercial Use
+
+"Commercial Use" refers to any use undertaken for direct or indirect economic gain, including use intended to generate operating revenue, service revenue, customer contract revenue, subscription fees, advertising revenue, commissions, license fees, or comparable commercial returns.
+
+### 1.9 Redistribution
+
+"Redistribution" means providing, sharing, transferring, selling, leasing, lending, mirroring, hosting, publicly releasing, exposing by API or download, or otherwise enabling a third party to obtain the Dataset or any substitutable part of it.
+
+### 1.10 Substantial Portion
+
+"Substantial Portion" means any content that, by quantity, quality, substitutability, reconstructability, or commercial value, is sufficient to substitute for the Dataset as a whole or for an important part of it.
+
+## 2. License Grant
+
+### 2.1 Limited License
+
+Subject to the Licensee's continuing compliance with this Agreement, OpenCSG grants the Licensee a limited, non-exclusive, non-transferable, non-sublicensable, revocable license.
+
+### 2.2 Scope of License
+
+This license permits the Licensee only to:
+
+1. access, download, and internally store the Dataset;
+2. use the Dataset for research, evaluation, validation, internal development, model training, model fine-tuning, retrieval augmentation, quality analysis, and compliance review;
+3. process the Dataset within infrastructure owned or controlled by the Licensee; and
+4. generate Results and Models based on the Dataset, subject to all restrictions in this Agreement relating to redistribution, commercialization, source constraints, and compliance obligations.
+
+### 2.3 Reserved Rights
+
+Except for the rights expressly granted under this Agreement, OpenCSG grants no other rights to the Licensee, whether express or implied.
+
+## 3. Source Rights and License Boundary
+
+### 3.1 Scope of Rights
+
+The Dataset may contain materials from public sources, procured sources, partner sources, or other third-party sources. The Licensee acknowledges that OpenCSG grants rights under this Agreement only to the extent that OpenCSG lawfully holds or may grant such rights.
+
+### 3.2 Priority of Specific Restrictions
+
+If any specific file, subset, field, sample, or source material within the Dataset carries stricter source notes, license terms, use restrictions, or compliance requirements, those stricter terms shall prevail.
+
+### 3.3 No Independent License to Source Materials
+
+This Agreement does not grant the Licensee any independent rights in the Source Materials themselves, nor any redistribution, relicensing, or sublicensing rights beyond OpenCSG's actual scope of rights.
+
+### 3.4 No Grant of Other Third-Party Rights
+
+The Licensee shall not claim under this Agreement any right to obtain any third-party trademark right, attribution right, moral right, database right, privacy right, data right, trade secret right, or any other right not expressly granted by OpenCSG.
+
+## 4. Use Restrictions
+
+### 4.1 Activities Requiring Prior Written Authorization
+
+Unless OpenCSG has provided prior written authorization, the Licensee shall not:
+
+1. redistribute the Dataset, Enhanced Data, or any Substantial Portion of the Dataset;
+2. provide the Dataset through mirrors, download sites, data packages, SaaS, PaaS, APIs, hosted retrieval services, data interfaces, or other external service methods;
+3. embed, bundle, or deliver the Dataset within commercial products, commercial services, data products, data services, or paid external solutions;
+4. sell, resell, relicense, lease, lend, or exchange the Dataset; or
+5. publish sample collections, split results, caches, indexes, or middleware that would allow a third party to reconstruct, extract, or replace the Dataset as a whole or any important part of it.
+
+### 4.2 Rights Management and Technical Controls
+
+The Licensee shall not:
+
+1. remove, obscure, alter, or falsify source information, attribution information, version information, license markings, source paths, copyright notices, or other rights management information in the Dataset;
+2. circumvent access controls, license controls, rate limits, download limits, feature gating, or security measures established by OpenCSG;
+3. use the Dataset to build data distribution services, data aggregation services, or data trading services that directly substitute for OpenCSG datasets; or
+4. state or imply that OpenCSG endorses the Licensee, the Licensee's products, the Licensee's services, or the Licensee's Results.
+
+### 4.3 Prohibited Unlawful Use
+
+The Licensee shall not use the Dataset for any unlawful, infringing, discriminatory, fraudulent, abusive, or otherwise non-compliant purpose.
+
+## 5. Results and Models
+
+### 5.1 Generation of Results and Models
+
+The Licensee may generate Results and Models based on the Dataset.
+
+### 5.2 No Circumvention Through Results or Models
+
+Results and Models are not automatically deemed to be the Dataset itself, but the Licensee shall not use Results or Models as a means to circumvent the restrictions in this Agreement applicable to the Dataset, Enhanced Data, or any Substantial Portion.
+
+### 5.3 No Release of Reconstructive Artifacts
+
+The Licensee shall not publish, deliver, or commercialize any Results, Models, weights, indexes, caches, vector stores, retrieval stores, or other artifacts that are sufficient to enable a third party to recover, extract, locate, reconstruct, or materially substitute for the contents of the Dataset at scale.
+
+### 5.4 Permitted Public Outputs
+
+The Licensee may publish papers, evaluation findings, statistical indicators, benchmark results, and limited non-substantial quotations, provided that reasonable attribution is retained and no substantial Dataset content is disclosed beyond what is necessary.
+
+### 5.5 Supplemental Authorization for Commercial Models and Systems
+
+If the Licensee intends to provide externally any model, system, Agent, API service, or commercial product trained, fine-tuned, distilled, or enhanced using the Dataset, OpenCSG may require a separate commercial license or supplemental agreement based on the data source, use scope, delivery method, and business model.
+
+## 6. Attribution and Notice Obligations
+
+### 6.1 Reasonable Attribution
+
+When the Licensee uses the Dataset in public citations, publications, product descriptions, evaluation reports, or external communications, the Licensee shall reasonably identify:
+
+1. the name of the Dataset;
+2. OpenCSG;
+3. the Dataset version;
+4. the Dataset link or the link to this Agreement; and
+5. any other attribution or source notice that OpenCSG requires to be retained.
+
+### 6.2 Notice of Modifications
+
+If the Licensee publicly releases Enhanced Data or non-substantial excerpts within the permitted scope, the Licensee shall clearly indicate that modifications were made and shall retain the original attribution, source notices, and license notices.
+
+### 6.3 Preservation of Notices
+
+The Licensee shall not remove or weaken OpenCSG's rights notices, disclaimers, liability limitations, source notices, or license notices.
+
+## 7. Compliance and Audit
+
+### 7.1 Compliance Obligations
+
+The Licensee shall ensure that its receipt, storage, processing, release, training, deployment, and commercialization activities involving the Dataset comply with applicable laws, regulations, and regulatory requirements, including without limitation those relating to intellectual property, data compliance, personal information protection, cybersecurity, export controls, and sector-specific regulation.
+
+### 7.2 Compliance Explanation
+
+If OpenCSG reasonably believes that the Licensee has engaged in non-compliant use, overbroad redistribution, unauthorized commercialization, circumvention of restrictions, or other material breach, OpenCSG may require the Licensee to provide the necessary compliance explanation.
+
+### 7.3 Audit Rights
+
+Upon reasonable notice, OpenCSG may audit, verify, or require compliance reporting with respect to the performance of material obligations under this Agreement.
+
+## 8. Intellectual Property
+
+### 8.1 Ownership
+
+All rights, title, and interest in the Dataset, its related documentation, metadata structures, version organization, identifier systems, descriptive text, and related materials provided by OpenCSG remain with OpenCSG or its lawful right holders.
+
+### 8.2 No Transfer of Rights
+
+Except for the limited rights of use expressly granted under this Agreement, no transfer of intellectual property rights occurs.
+
+### 8.3 Trademark Protection
+
+OpenCSG, CSGHub, and related names, logos, trademarks, service marks, and brand elements are protected by law and may not be used without prior written permission.
+
+## 9. Disclaimer
+
+### 9.1 As-Is Basis
+
+The Dataset is provided on an "AS IS" basis.
+
+### 9.2 No Warranties
+
+To the maximum extent permitted by applicable law, OpenCSG makes no express or implied warranties regarding the Dataset, including without limitation as to title, completeness, continued availability, accuracy, legality, merchantability, fitness for a particular purpose, non-infringement, absence of errors, absence of interruption, or absence of security defects.
+
+### 9.3 Assumption of Risk
+
+The Licensee shall determine whether the Dataset is suitable for its specific purposes and shall bear the related risks.
+
+## 10. Limitation of Liability
+
+### 10.1 Indirect Damages
+
+To the maximum extent permitted by applicable law, OpenCSG shall not be liable for any indirect, incidental, special, punitive, or consequential damages.
+
+### 10.2 Business and Data Losses
+
+OpenCSG shall not be liable for loss of business, loss of profits, loss of goodwill, loss of opportunity, loss of data, loss of models, interruption losses, or third-party claims.
+
+### 10.3 Scope of Liability
+
+Whether based on contract, tort, strict liability, or any other legal theory, OpenCSG's liability under this Agreement shall be limited to the minimum extent permitted by applicable law.
+
+## 11. Termination
+
+### 11.1 Grounds for Termination
+
+If the Licensee breaches any provision of this Agreement, or engages in unauthorized redistribution, commercialization, circumvention of restrictions, infringing use, or other material breach, the license under this Agreement may be terminated by OpenCSG immediately or terminate automatically.
+
+### 11.2 Obligations Upon Termination
+
+Upon termination, the Licensee shall immediately:
+
+1. cease receiving, using, processing, and externally providing the Dataset;
+2. delete or destroy the Dataset and any unauthorized retained copies under its control; and
+3. provide written confirmation of deletion or cessation of use upon OpenCSG's request.
+
+### 11.3 Materials That May Be Retained
+
+Results, papers, and compliance retention records lawfully created before termination and not containing any Substantial Portion of the Dataset may continue to be retained, provided that no other provision of this Agreement is violated.
+
+## 12. Commercial Authorization
+
+### 12.1 Scenarios Requiring Separate Authorization
+
+For any of the following scenarios, the Licensee shall contact OpenCSG in advance to obtain a separate commercial license, OEM license, distribution license, or supplemental authorization:
+
+1. redistribution of the Dataset;
+2. provision to third parties of Dataset download, mirroring, hosting, retrieval, invocation, or access services;
+3. embedding the Dataset into commercial products, commercial services, SaaS, PaaS, APIs, or paid offerings;
+4. external provision of models, Agents, services, or solutions trained, fine-tuned, distilled, enhanced, or materially supported by the Dataset; or
+5. resale, bundling, subcontracting, white-labeling, OEM distribution, channel distribution, or delivery of industry solutions.
+
+### 12.2 Contact
+
+Commercial licensing contact:
+lorraineg@opencsg.com
+
+## 13. Governing Law and Dispute Resolution
+
+### 13.1 Governing Law
+
+The formation, validity, interpretation, performance, and dispute resolution of this Agreement shall be governed by the law of the place where the OpenCSG-designated entity is located; the default may be the laws of the People's Republic of China.
+
+### 13.2 Optional Jurisdictions
+
+If required by OpenCSG's group structure or transaction arrangements, the formal version may instead designate Singapore, the Cayman Islands, or another appropriate jurisdiction.
+
+### 13.3 Dispute Resolution
+
+Any dispute arising out of or in connection with this Agreement shall first be resolved through friendly consultation. If consultation fails, the dispute shall be submitted to the court or arbitral institution designated in the formal version of this Agreement.
+
+## 14. Miscellaneous
+
+### 14.1 Updates to the Agreement
+
+OpenCSG may publish updated versions of this Agreement within a reasonable scope. Unless otherwise specified, the Licensee's continued use of the Dataset constitutes acceptance of the new terms applicable to subsequent access or subsequent versions.
+
+### 14.2 Severability
+
+If any provision of this Agreement is held invalid or unenforceable, the validity and enforceability of the remaining provisions shall not be affected.
+
+### 14.3 Entire Agreement
+
+This Agreement constitutes the entire agreement between the Licensee and OpenCSG regarding the use of the Dataset.

README.md

@@ -0,0 +1,600 @@
+---
+pretty_name: CIMD
+license: other
+language:
+  - zh
+  - en
+language_creators:
+  - found
+multilinguality:
+  - multilingual
+source_datasets:
+  - original
+size_categories:
+  - 100K<n<1M
+tags:
+  - text
+  - jsonl
+  - mining
+  - iron-ore
+  - metallurgy
+  - law
+  - finance
+  - industry
+  - rag
+  - domain-dataset
+configs:
+  - config_name: state_laws
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/state_laws/train.jsonl"
+  - config_name: domestic_conference_papers
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/domestic_conference_papers/train.jsonl"
+  - config_name: doctoral_dissertations
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/doctoral_dissertations/train.jsonl"
+  - config_name: chinese_journals
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/chinese_journals/train.jsonl"
+  - config_name: international_journal_of_mining_science_and_technology
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/international_journal_of_mining_science_and_technology/train.jsonl"
+  - config_name: international_journal_of_minerals_metallurgy_and_materials
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/international_journal_of_minerals_metallurgy_and_materials/train.jsonl"
+  - config_name: industry_research_reports
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/industry_research_reports/train.jsonl"
+  - config_name: sintering_and_pelletizing
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/sintering_and_pelletizing/train.jsonl"
+  - config_name: ironmaking
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/ironmaking/train.jsonl"
+  - config_name: public_opinion
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/public_opinion/train.jsonl"
+  - config_name: books
+    data_files:
+      - split: train
+        path: "data/text/iron_ore/books/train.jsonl"
+---
+
+# CIMD
+
+<div align="center">
+    <a href="#%E4%B8%AD%E6%96%87%E8%AF%B4%E6%98%8E" target="_self">[[中文]]</a> | <a href="#english" target="_self">[[English]]</a>
+</div>
+
+<div align="center">
+
+[CSGHub](https://opencsg.com/datasets/OpenCSG/CIMD?tab=summary) | [Hugging Face](https://huggingface.co/datasets/opencsg/CIMD) | [OpenCSG Community](https://opencsg.com/models) | [👾 GitHub](https://github.com/OpenCSGs/csghub)
+
+</div>
+
+## 中文说明
+
+### 数据集概述
+
+CIMD 是面向铁矿石及矿冶产业链的跨来源文本数据集。数据覆盖法律法规、行业规章、政策文件、行业标准、学术论文、会议论文、博士论文、图书、行业研究、企业经营、产能产量、市场交易与舆情观点等核心类型，当前公开版本以统一的 JSONL 记录格式发布，并保留标题、来源类型、作者、时间、语言、关键词、授权类型与来源信息等关键元数据字段。
+
+![CIMD 概览](./imgs/02_intro.png)
+
+**跨来源整合。** CIMD 将制度文本、技术文本、研究文本、经营文本和市场文本放入同一数据体系。单一来源的数据通常只能回答局部问题，例如政策库侧重制度依据，论文库侧重技术原理，市场库侧重价格和交易信号，而行业实际问题往往同时依赖法规边界、工艺机理、企业行为和市场环境。CIMD 将法律法规、政策文件、行业标准、学术论文、博士论文、图书、行业研究、企业经营信息、产能产量、市场交易与舆情观点统一组织，使同一主题可以在多种来源之间形成连续证据链。对于行业检索、RAG、Agent 推理、专题研究、领域微调和评测集构建，这种跨来源结构能够更直接地支撑“制度依据 + 技术材料 + 市场证据”的联合调用，减少跨库拼接带来的语义割裂和上下文缺口。
+
+![CIMD 跨来源结构](./imgs/04_cross_source.png)
+
+**权威来源支撑。** CIMD 汇集了矿冶行业中具有代表性的权威文本。制度层面，数据覆盖国家法律法规、行业规章、政策文件和标准文本，适合构建具备制度依据和合规边界的行业知识底座。学术层面，数据覆盖中文期刊、国内会议论文、博士学位论文，并纳入 `International Journal of Mining Science and Technology`、`International Journal of Minerals, Metallurgy and Materials` 等国际矿冶期刊内容，形成兼具中文产业语境与国际技术视野的专业语料。产业层面，数据进一步延伸至科研院所报告、行业协会报告、券商研究、企业经营信息、产能产量与市场相关材料。来源主体稳定、专业、可追溯，使数据在专业深度、行业可信度和实际应用价值上具备更高上限。
+
+**完整数据体系。** CIMD 的数据组织不是简单的文件汇总，而是围绕铁矿石及矿冶产业链建立的行业数据体系。完整体系覆盖 `9` 个一级分类、`42` 个二级分类、`335` 个三级/四级来���节点，涵盖法律文件与法规依据、行业规章与管理办法、政策文件与行业指导、行业标准、专利与知识产权、学术与培训资料、互联网舆情与观点分析、企业经营与运营信息、行业研究与市场报告等核心门类。公开快照围绕铁矿石资源、采选加工、烧结球团、炼铁生产、政策治理、企业经营与市场观察形成连续语料结构，覆盖制度依据、技术机理、生产流程、经营信息和市场信号。该体系提升数据覆盖深度，并为专题扩展、增量采集、行业补数、基准任务设计和数据资产编目提供结构框架。
+
+**元数据完整。** 当前版本在每条记录中保留 `file_id`、`data_id`、`title`、`source_type`、`author`、`original_time`、`content_time`、`language`、`keywords`、`license_type`、`source_details` 等关键字段。使用者可以按来源、时间、语言、主题和授权边界直接筛选样本，也可以把检索到的文本片段回溯到原始文件和具体记录。对于长文档检索、来源归因、审计留痕、授权控制、质量抽检和数据资产管理，这类记录级元数据比单纯正文更有操作价值，也更适合持续扩展和版本化维护。
+
+**面向标准化建设与可信流通。** 数据集把标识、来源、分类、时间、授权和来源说明直接放在记录体内，而不是把这些信息留在外部台账里。这样一来，同一份快照可以直接进入数据目录编制、质量抽检、授权审计、责任追踪和可信流通流程，不需要在语料之外再补一套独立的元数据体系。这种组织方式与国家数据局和全国数据标准化技术委员会公开发布的《高质量数据集 建设指南》《高质量数据集 格式要求》《高质量数据集 分类指南》《高质量数据集 质量评测规范》以及《国家数据基础设施建设指引》强调的方向保持同向。
+
+**直接可用于模型与应用。** 当前公开版本以统一 JSONL 记录格式发布，不是单纯的原始 PDF 堆积。解析后的记录可直接进入检索、切分、标注、训练、评测和服务流程，便于快速接入 CSGHub / OpenCSG 数据平台，以及 Hugging Face、ModelScope、自建 RAG 系统和行业智能体工作流。对于需要高质量中文专业语料、制度与技术联合语料、跨来源证据检索能力和可管理元数据结构的行业智能场景，CIMD 具备较强的即用性。
+
+当前公开版本包含 11 个子集，共 382,425 条有效 JSONL 记录，对应 57,622 个去重 `file_id`。完整数据体系覆盖 `9` 个一级分类、`42` 个二级分类、`335` 个三级/四级来源节点。
+
+![CIMD 数据规模](./imgs/08_data_scale.png)
+
+公开仓库地址：
+
+- [CSGHub / OpenCSG / CIMD](https://opencsg.com/datasets/OpenCSG/CIMD?tab=summary)
+- [Hugging Face / opencsg / CIMD](https://huggingface.co/datasets/opencsg/CIMD)
+- `https://opencsg.com/datasets/OpenCSG/CIMD.git`
+
+### 标准与参考文件
+
+- [国家数据局：第一批高质量数据集方向标准/技术文件验证试点典型单位发布](https://www.nda.gov.cn/sjj/ywpd/sjzg/0922/20250831180245645997474_pc.html)
+- [全国数据标准化技术委员会：关于发布高质量数据集、全国一体化算力网、数据基础设施、可信数据空间等4个方向19项技术文件的通知](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b.shtml)
+- [技术文件《高质量数据集 建设指南》](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/1.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E5%BB%BA%E8%AE%BE%E6%8C%87%E5%8D%97%E3%80%8B.pdf.pdf)
+- [技术文件《高质量数据集 格式要求》](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/2.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E6%A0%BC%E5%BC%8F%E8%A6%81%E6%B1%82%E3%80%8B.pdf.pdf)
+- [技术文件《高质量数据集 分类指南》](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/3.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E5%88%86%E7%B1%BB%E6%8C%87%E5%8D%97%E3%80%8B.pdf.pdf)
+- [技术文件《高质量数据集 质量评测规范》](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/4.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E8%B4%A8%E9%87%8F%E8%AF%84%E6%B5%8B%E8%A7%84%E8%8C%83%E3%80%8B.pdf.pdf)
+- [国家数据基础设施建设指引](https://www.nda.gov.cn/form-files-sjj/2c99e457-8f9edeed-018f-b8a7537c-0243/2c99e457-8f9edeed-018f-b8ad7daf-0256/fileupload_scfj/2024/11/22/ff808081-9348abb8-0193-51e366ec-1784.pdf?id=ff808081-9348abb8-0193-51e366ec-1784)
+
+### 数据组成
+
+当前公开快照按子集组织在 `data/text/iron_ore/` 目录下。
+
+| 子集 | JSONL 记录��数 | 去重 `file_id` 数 | 存储方式 | 内容 |
+| --- | ---: | ---: | --- | --- |
+| `state_laws` | 99,496 | 7,300 | Git LFS | 法律法规、规章制度、政策文本 |
+| `domestic_conference_papers` | 58,221 | 18,826 | Git LFS | 国内会议论文与会议资料 |
+| `doctoral_dissertations` | 37,961 | 804 | Git LFS | 博士学位论文 |
+| `public_opinion` | 30,705 | 9,427 | Git LFS | 舆情与观点资料 |
+| `chinese_journals` | 28,266 | 6,412 | Git LFS | 中文期刊论文 |
+| `international_journal_of_mining_science_and_technology` | 16,824 | 2,435 | Git LFS | 英文学术期刊 |
+| `international_journal_of_minerals_metallurgy_and_materials` | 15,391 | 2,461 | Git LFS | 英文学术期刊 |
+| `industry_research_reports` | 14,319 | 1,182 | Git LFS | 行业研究、券商、企业与产能相关材料 |
+| `sintering_and_pelletizing` | 9,558 | 3,783 | Git LFS | 烧结球团专题资料 |
+| `ironmaking` | 8,002 | 3,871 | Git LFS | 炼铁与生产专题资料 |
+| `books` | 63,682 | 1,121 | Git LFS | 图书资料 |
+
+### 快照统计
+
+以下统计基于 `2026-03-25` 的仓库快照。
+
+| 项目 | 数值 |
+| --- | --- |
+| 已声明子集配置 | 11 |
+| 当前有效 JSONL 记录条数 | 382,425 |
+| 去重 `file_id` 数 | 57,622 |
+| `source_type` 类别数 | 52 |
+| 数据体系层级 | 9 个一级分类，42 个二级分类，335 个三级/四级来源节点 |
+
+统计口径：
+
+- `JSONL 记录条数` 指有效 JSONL 记录数，不是原始文件数。
+- 统计时已排除空行、占位行和无效首行。
+- 一份原始文件可以拆分为多条 JSONL 记录。
+- `去重 file_id 数` 用于表示子集内可识别的源文件数量。
+- `Git LFS` 表示文件内容由 Git Large File Storage 管理；普通 `git clone` 之后如需拉取实际内容，需要执行 `git lfs pull`。
+
+语言分布（按 JSONL 记录条数）：
+
+| 语言 | 记录数 |
+| --- | ---: |
+| `zh` | 232,169 |
+| `en` | 115,113 |
+| `other` | 35,143 |
+
+`format` 字段分布（按 JSONL 记录条数）：
+
+| 格式 | 记录数 |
+| --- | ---: |
+| `pdf` | 349,410 |
+| `jsonl` | 30,705 |
+| `docx` | 2,054 |
+| `markdown` | 249 |
+| `doc` | 7 |
+
+主要来源类型（按 JSONL 记录条数）：
+
+| `source_type` | 记录数 |
+| --- | ---: |
+| 期刊论文 | 116,926 |
+| 国家法律法规 | 95,394 |
+| 学术出版物 | 57,492 |
+| 学位论文 | 38,075 |
+| 社会公众与自媒体舆情 | 31,178 |
+| 企业基本信息 | 11,656 |
+| 产能与产量数据 | 6,282 |
+| 科研院所报告 | 5,143 |
+| 行业协会报告 | 4,664 |
+| 国内产业政策 | 2,393 |
+| 券商与投行报告 | 2,343 |
+| 期货衍生品交易数据 | 1,405 |
+
+### 数据体系
+
+完整数据体系按以下九类组织：
+
+1. 法律文件与法规依据
+2. 行业规章与管理办法
+3. 政策文件与行业指导
+4. 行业标准
+5. 专利与知识产权
+6. 学术与培训资料
+7. 互联网舆情与观点分析
+8. 企业经营与运营信息
+9. 行业研究与市场报告
+
+### 目录结构
+
+```text
+data/
+  text/
+    iron_ore/
+      state_laws/train.jsonl
+      domestic_conference_papers/train.jsonl
+      doctoral_dissertations/train.jsonl
+      chinese_journals/train.jsonl
+      international_journal_of_mining_science_and_technology/train.jsonl
+      international_journal_of_minerals_metallurgy_and_materials/train.jsonl
+      industry_research_reports/train.jsonl
+      sintering_and_pelletizing/train.jsonl
+      ironmaking/train.jsonl
+      public_opinion/train.jsonl
+      books/train.jsonl
+```
+
+### 数据结构
+
+当前仓库以 JSONL 为主，每行对应一条解析记录。单个源文件可以对应多条记录，因此文件数与记录数不是同一统计口径。
+
+| 字段 | 类型 | 说明 |
+| --- | --- | --- |
+| `format` | string | 来源文件或载体格式 |
+| `file_id` | string | 文件标识 |
+| `raw_chunk` | string | 解析后的文本内容 |
+| `file_name` | string | 原始文件名 |
+| `title` | string | 标题 |
+| `source_type` | string | 来源类型 |
+| `author` | string | 作者、机构或发布主体 |
+| `original_time` | string | 原始发布时间 |
+| `content_time` | string | 内容时间 |
+| `source_details` | string | 公开来源链接或来源说明 |
+| `data_version` | string | 数据版本 |
+| `license_type` | string | 记录级授权类型 |
+| `is_generated` | string | 是否标记为生成内容 |
+| `country` | string | 国家标签 |
+| `language` | string | 语言标签 |
+| `keywords` | array | 关键词 |
+| `data_id` | string | 记录标识 |
+
+示例：
+
+```json
+{
+  "format": "pdf",
+  "file_id": "241f81dd17a546245faf672cf36fcecd",
+  "file_name": "iron_ore_report.pdf",
+  "data_id": "241f81dd17a546245faf672cf36fcecd",
+  "title": "铁矿石价格（1月3日）",
+  "source_type": "科研院所报告",
+  "author": "中国矿产资源研究院",
+  "original_time": "2024-01-04 00:00:00",
+  "content_time": "2024-01-03 00:00:00",
+  "data_version": "1.0.0",
+  "is_generated": "0",
+  "country": "中国",
+  "language": "zh",
+  "keywords": [
+    "现货交易",
+    "金融衍生品"
+  ],
+  "license_type": "商业授权",
+  "raw_chunk": "...",
+  "source_details": "https://example.com/report/iron-ore-2024-01-03"
+}
+```
+
+### 获取与加载
+
+通过 Git 获取：
+
+```bash
+git lfs install
+git clone https://opencsg.com/datasets/OpenCSG/CIMD.git
+cd CIMD
+git lfs pull
+```
+
+使用 Hugging Face `datasets`：
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset(
+    "opencsg/CIMD",
+    "state_laws",
+    split="train",
+    streaming=True,
+)
+```
+
+使用 ModelScope：
+
+```python
+from modelscope.msdatasets import MsDataset
+
+dataset = MsDataset.load(
+    dataset_name="CIMD",
+    namespace="OpenCSG",
+    subset_name="state_laws",
+    split="train",
+)
+```
+
+### 适用场景
+
+- 铁矿石与矿冶行业垂直检索与 RAG
+- 行业研究助手与证据检索系统
+- 政策、法规与合规问答
+- 长文档问答与跨来源证据聚合
+- 领域继续预训练、SFT 样本构建与数据筛选
+- 文档分类、来源识别、主题标注与术语抽取
+
+### 使用注意
+
+- 当前计数单位为解析记录，不等同于去重后的原始文档数。
+- 当前公开子集通过 Git LFS 管理。
+- 不同来源之间可能存在重复、近重复或解析噪声。
+- 时间字段可能表示发布时间、内容时间或抽取时间。
+- 用于训练、分发或商用前，需要结合来源信息核验实际授权范围。
+
+### 许可说明
+
+使用本数据集需要遵循 [OpenCSG 数据集社区许可协议](./OpenCSG数据集社区许可协议.md)。仓库级 metadata 使用 `license: other`，当前公开快照中的记录级 `license_type` 统一为 `商业授权`。本数据集支持商业用途。如计划对外提供本数据集、基于本数据集训练或增强的模型、系统、Agent、API 服务或商业产品，请遵守该许可协议，并发送邮件至 `lorraineg@opencsg.com` 获取商业授权。
+
+### 引用
+
+```bibtex
+@dataset{opencsg_cimd_2026,
+  title        = {CIMD: A Cross-Source Industry Corpus for Iron Ore, Mining, Metallurgy, Policy, and Market Intelligence},
+  author       = {OpenCSG},
+  year         = {2026},
+  url          = {https://opencsg.com/datasets/OpenCSG/CIMD},
+  note         = {OpenCSG dataset repository}
+}
+```
+
+---
+
+<div align="center">
+    <a href="#%E4%B8%AD%E6%96%87%E8%AF%B4%E6%98%8E" target="_self">[[中文]]</a> | <a href="#english" target="_self">[[English]]</a>
+</div>
+
+## English
+
+### Overview
+
+CIMD is a cross-source text dataset for iron ore and the broader mining-metallurgy value chain. The current public release provides normalized JSONL records that cover legal and policy documents, academic literature, conference materials, doctoral dissertations, books, industry research, enterprise information, production data, market-related content, and public opinion materials. The repository preserves per-record metadata including title, source type, author, timestamps, language, keywords, license tags, and source information.
+
+![CIMD overview](./imgs/01_cover_en.png)
+
+**Cross-source integration.** CIMD brings regulatory, technical, research, operational, and market texts into one data system. Single-source collections answer one part of an industry question. Policy repositories provide regulatory boundaries, academic repositories provide technical mechanisms, and market repositories provide transaction and sentiment signals. CIMD organizes laws and regulations, policy documents, standards, academic papers, doctoral dissertations, books, industry research, enterprise information, production data, market materials, and public opinion into one corpus, so users can track the same topic across source types. This structure fits retrieval, evidence aggregation, agent workflows, domain tuning, and benchmark construction.
+
+![CIMD cross-source structure](./imgs/02_cross_source_en.png)
+
+**Authoritative sources.** CIMD collects representative and traceable texts from the mining-metallurgy domain. On the institutional side, it covers national laws and regulations, industry rules, policy documents, and standards. On the academic side, it covers `zh` journal collections, domestic conference papers, doctoral dissertations, and English technical literature from journals such as `International Journal of Mining Science and Technology` and `International Journal of Minerals, Metallurgy and Materials`. On the industry side, it extends to research institute reports, industry association reports, brokerage research, enterprise information, production data, and market-facing materials. This source composition gives the dataset strong domain credibility and practical value.
+
+**Comprehensive taxonomy.** CIMD uses an industry data system rather than a flat document dump. The full structure covers `9` first-level categories, `42` second-level categories, and `335` third/fourth-level source nodes, including legal and regulatory materials, administrative rules, policy guidance, standards, patents and intellectual property, academic and training materials, public opinion and commentary, enterprise operations, and industry research and market reports. The public snapshot forms a continuous corpus around iron ore resources, beneficiation and processing, sintering and pelletizing, ironmaking, policy governance, enterprise operations, and market observation. This structure supports expansion, incremental collection, data cataloging, and benchmark design.
+
+**Rich metadata.** Each record keeps fields such as `file_id`, `data_id`, `title`, `source_type`, `author`, `original_time`, `content_time`, `language`, `keywords`, `license_type`, and `source_details`. Users can filter by source, time window, language, topic, or authorization boundary, and they can trace a retrieved chunk back to the underlying record and source file. For long-document retrieval, provenance tracking, audit logs, license control, and ongoing dataset maintenance, this is materially more useful than plain text alone.
+
+**Alignment with standardization and trusted circulation.** The dataset keeps identifiers, source labels, classification fields, timestamps, authorization tags, and source notes inside each record instead of leaving them in a separate ledger. That means the same snapshot can move into cataloging, quality review, authorization checks, accountability workflows, and trusted data-space services without a second metadata pass outside the corpus. This layout follows the direction reflected in the published `High-Quality Dataset Construction Guide`, `High-Quality Dataset Format Requirements`, `High-Quality Dataset Classification Guide`, `High-Quality Dataset Quality Evaluation Specification`, and the `Guidelines for National Data Infrastructure Development`.
+
+**Ready for models and applications.** The public release uses normalized JSONL records rather than raw source files alone. The parsed records move directly into retrieval, chunking, annotation, training, evaluation, and serving workflows, and teams can plug them into the CSGHub / OpenCSG data platform, Hugging Face, ModelScope, custom RAG systems, and industry agent pipelines. For domain applications that require `zh` materials, combined regulatory and technical evidence, cross-source retrieval, and manageable metadata, CIMD is directly usable.
+
+The current public release contains 11 subsets, 382,425 effective JSONL records, and 57,622 unique `file_id` values. The full taxonomy covers 9 first-level categories, 42 second-level categories, and 335 third/fourth-level source nodes.
+
+![CIMD data scale](./imgs/03_data_scale_en.png)
+
+Public repository:
+
+- [CSGHub / OpenCSG / CIMD](https://opencsg.com/datasets/OpenCSG/CIMD?tab=summary)
+- [Hugging Face / opencsg / CIMD](https://huggingface.co/datasets/opencsg/CIMD)
+- `https://opencsg.com/datasets/OpenCSG/CIMD.git`
+
+### Standards and References
+
+- [National Data Administration: Release of the first batch of pilot units for high-quality dataset standards and technical documents](https://www.nda.gov.cn/sjj/ywpd/sjzg/0922/20250831180245645997474_pc.html)
+- [National Data Standardization Technical Committee: Notice on releasing 19 technical documents across high-quality datasets, the national integrated computing network, data infrastructure, and trusted data spaces](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b.shtml)
+- [High-Quality Dataset Construction Guide](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/1.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E5%BB%BA%E8%AE%BE%E6%8C%87%E5%8D%97%E3%80%8B.pdf.pdf)
+- [High-Quality Dataset Format Requirements](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/2.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E6%A0%BC%E5%BC%8F%E8%A6%81%E6%B1%82%E3%80%8B.pdf.pdf)
+- [High-Quality Dataset Classification Guide](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/3.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E5%88%86%E7%B1%BB%E6%8C%87%E5%8D%97%E3%80%8B.pdf.pdf)
+- [High-Quality Dataset Quality Evaluation Specification](https://www.tc609.org.cn/tc609/tzgg/202509/6d0f135392ca47dfa28dad22f7bb5f6b/files/4.%20%E6%8A%80%E6%9C%AF%E6%96%87%E4%BB%B6%E3%80%8A%E9%AB%98%E8%B4%A8%E9%87%8F%E6%95%B0%E6%8D%AE%E9%9B%86%20%E8%B4%A8%E9%87%8F%E8%AF%84%E6%B5%8B%E8%A7%84%E8%8C%83%E3%80%8B.pdf.pdf)
+- [Guidelines for National Data Infrastructure Development](https://www.nda.gov.cn/form-files-sjj/2c99e457-8f9edeed-018f-b8a7537c-0243/2c99e457-8f9edeed-018f-b8ad7daf-0256/fileupload_scfj/2024/11/22/ff808081-9348abb8-0193-51e366ec-1784.pdf?id=ff808081-9348abb8-0193-51e366ec-1784)
+
+### Data Composition
+
+The current public snapshot sits under `data/text/iron_ore/`.
+
+| Subset | JSONL records | Unique `file_id` values | Storage | Content |
+| --- | ---: | ---: | --- | --- |
+| `state_laws` | 99,496 | 7,300 | Git LFS | Laws, regulations, policy, and institutional texts |
+| `domestic_conference_papers` | 58,221 | 18,826 | Git LFS | Domestic conference papers and conference materials |
+| `doctoral_dissertations` | 37,961 | 804 | Git LFS | Doctoral dissertations |
+| `public_opinion` | 30,705 | 9,427 | Git LFS | Public opinion and commentary materials |
+| `chinese_journals` | 28,266 | 6,412 | Git LFS | Chinese academic journals |
+| `international_journal_of_mining_science_and_technology` | 16,824 | 2,435 | Git LFS | English academic literature |
+| `international_journal_of_minerals_metallurgy_and_materials` | 15,391 | 2,461 | Git LFS | English academic literature |
+| `industry_research_reports` | 14,319 | 1,182 | Git LFS | Industry research, brokerage, enterprise, and production-related materials |
+| `sintering_and_pelletizing` | 9,558 | 3,783 | Git LFS | Sintering and pelletizing materials |
+| `ironmaking` | 8,002 | 3,871 | Git LFS | Ironmaking and production materials |
+| `books` | 63,682 | 1,121 | Git LFS | Books |
+
+### Snapshot Statistics
+
+The following statistics correspond to the repository snapshot dated `2026-03-25`.
+
+| Item | Value |
+| --- | --- |
+| Declared subset configs | 11 |
+| Effective JSONL records | 382,425 |
+| Unique `file_id` values | 57,622 |
+| Distinct `source_type` values | 52 |
+| Taxonomy depth | 9 first-level categories, 42 second-level categories, 335 third/fourth-level source nodes |
+
+Counting rules:
+
+- `JSONL records` refers to effective JSONL records, not source-file counts.
+- The counts exclude empty lines, placeholder lines, and non-record first lines.
+- One source file can produce more than one JSONL record.
+- `Unique file_id values` gives the file-level count within a subset.
+- `Git LFS` means Git Large File Storage holds the file content, and users may need to run `git lfs pull` after cloning.
+
+Language distribution by JSONL records:
+
+| Language | Records |
+| --- | ---: |
+| `zh` | 232,169 |
+| `en` | 115,113 |
+| `other` | 35,143 |
+
+`format` field distribution by JSONL records:
+
+| Format | Records |
+| --- | ---: |
+| `pdf` | 349,410 |
+| `jsonl` | 30,705 |
+| `docx` | 2,054 |
+| `markdown` | 249 |
+| `doc` | 7 |
+
+Dominant source types by JSONL records:
+
+| `source_type` | Records |
+| --- | ---: |
+| Journal papers | 116,926 |
+| National laws and regulations | 95,394 |
+| Academic publications | 57,492 |
+| Dissertations | 38,075 |
+| Public opinion and self-media content | 31,178 |
+| Enterprise basic information | 11,656 |
+| Production and capacity data | 6,282 |
+| Research institute reports | 5,143 |
+| Industry association reports | 4,664 |
+| Domestic industrial policy | 2,393 |
+| Brokerage and investment bank reports | 2,343 |
+| Futures and derivatives trading data | 1,405 |
+
+### Taxonomy
+
+The full data system covers the following nine first-level categories:
+
+1. Legal documents and regulatory basis
+2. Industry rules and administrative measures
+3. Policy documents and industry guidance
+4. Industry standards
+5. Patents and intellectual property
+6. Academic and training materials
+7. Public opinion and commentary
+8. Enterprise operations and business information
+9. Industry research and market reports
+
+### Repository Layout
+
+```text
+data/
+  text/
+    iron_ore/
+      state_laws/train.jsonl
+      domestic_conference_papers/train.jsonl
+      doctoral_dissertations/train.jsonl
+      chinese_journals/train.jsonl
+      international_journal_of_mining_science_and_technology/train.jsonl
+      international_journal_of_minerals_metallurgy_and_materials/train.jsonl
+      industry_research_reports/train.jsonl
+      sintering_and_pelletizing/train.jsonl
+      ironmaking/train.jsonl
+      public_opinion/train.jsonl
+      books/train.jsonl
+```
+
+### Schema
+
+The repository primarily stores JSONL records. Each line corresponds to one parsed record.
+
+| Field | Type | Description |
+| --- | --- | --- |
+| `format` | string | Source file or carrier format |
+| `file_id` | string | File identifier |
+| `raw_chunk` | string | Parsed text content |
+| `file_name` | string | Original file name |
+| `title` | string | Title |
+| `source_type` | string | Source type |
+| `author` | string | Author, institution, or publishing body |
+| `original_time` | string | Original publication time |
+| `content_time` | string | Content time |
+| `source_details` | string | Public source URL or source note |
+| `data_version` | string | Record version |
+| `license_type` | string | Record-level license classification |
+| `is_generated` | string | Generated-content flag |
+| `country` | string | Country tag |
+| `language` | string | Language tag |
+| `keywords` | array | Keywords |
+| `data_id` | string | Record identifier |
+
+### Access and Loading
+
+Access via Git:
+
+```bash
+git lfs install
+git clone https://opencsg.com/datasets/OpenCSG/CIMD.git
+cd CIMD
+git lfs pull
+```
+
+Load with Hugging Face `datasets`:
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset(
+    "opencsg/CIMD",
+    "state_laws",
+    split="train",
+    streaming=True,
+)
+```
+
+Load with ModelScope:
+
+```python
+from modelscope.msdatasets import MsDataset
+
+dataset = MsDataset.load(
+    dataset_name="CIMD",
+    namespace="OpenCSG",
+    subset_name="state_laws",
+    split="train",
+)
+```
+
+### Intended Uses
+
+- Vertical retrieval and RAG for iron ore, mining, and metallurgy
+- Industry research assistants and evidence retrieval systems
+- Policy, regulation, and compliance QA
+- Long-document QA and cross-source evidence aggregation
+- Continued pretraining, SFT data construction, and corpus filtering
+- Document classification, source identification, topic tagging, and terminology extraction
+
+### Limitations
+
+- Counts in this card use parsed records rather than deduplicated source documents.
+- The public subsets rely on Git LFS.
+- Duplicate, near-duplicate, and parsing-noise cases may exist.
+- Time fields may reflect publication time, content time, or extraction artifacts.
+- Verify provenance and the effective authorization scope before redistribution or downstream commercial use.
+
+### License
+
+Use of this dataset is subject to the [OpenCSG Dataset Community License](./OpenCSG数据集社区许可协议.md). The repository-level metadata uses `license: other`, and the current public snapshot uses `商业授权` as the record-level `license_type`. The dataset supports commercial use. If you plan to provide this dataset, or any model, system, Agent, API service, or commercial product trained or enhanced with this dataset, you must comply with that license and contact `lorraineg@opencsg.com` for commercial authorization.
+
+### Citation
+
+```bibtex
+@dataset{opencsg_cimd_2026,
+  title        = {CIMD: A Cross-Source Industry Corpus for Iron Ore, Mining, Metallurgy, Policy, and Market Intelligence},
+  author       = {OpenCSG},
+  year         = {2026},
+  url          = {https://opencsg.com/datasets/OpenCSG/CIMD},
+  note         = {OpenCSG dataset repository}
+}
+```

data/text/iron_ore/README.md

@@ -0,0 +1,32 @@
+# Iron Ore Text Collections
+
+## Layout
+
+All subsets use the following layout:
+
+```text
+data/text/iron_ore/<subset_name>/train.jsonl
+```
+
+Subset names use lowercase ASCII `snake_case`.
+
+## Subsets
+
+| Subset | Description |
+| --- | --- |
+| `books` | Books and long-form monographs |
+| `chinese_journals` | Chinese journal papers |
+| `doctoral_dissertations` | Doctoral dissertations |
+| `domestic_conference_papers` | Domestic conference papers and conference materials |
+| `industry_research_reports` | Industry, brokerage, and research reports |
+| `international_journal_of_minerals_metallurgy_and_materials` | International Journal of Minerals, Metallurgy and Materials |
+| `international_journal_of_mining_science_and_technology` | International Journal of Mining Science and Technology |
+| `ironmaking` | Ironmaking-related technical and production materials |
+| `public_opinion` | Public opinion and commentary materials |
+| `sintering_and_pelletizing` | Sintering and pelletizing materials |
+| `state_laws` | Laws, regulations, and institutional texts |
+
+## Notes
+
+- Public subset files use Git LFS.
+- The root [README.md](../../../README.md) contains JSONL record counts and unique `file_id` counts.

data/text/iron_ore/books/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0c4c85a02131bb7b7f05f5a8e5154bd29c0645ab9b78bb3158fbaa0c9e18a57
+size 629823759

data/text/iron_ore/chinese_journals/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af59f6da65c0dc721106d28066d94e707cf932f641908e6f6acde2958ffca711
+size 166484376

data/text/iron_ore/doctoral_dissertations/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a7507361996a1fbbde1984379d32d4b35b9bceccbe07d96a4012a07876891e4
+size 289892244

data/text/iron_ore/domestic_conference_papers/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4edd1c8d88b9f019bf1b610a7d028872528732c0922b38cb23be495994eeb95
+size 363073422

data/text/iron_ore/industry_research_reports/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:964ac3aee9c3cdb49ef07d45ae6deb6b31044038bd91d2f870b0c71a0e5ee6e5
+size 65527679

data/text/iron_ore/international_journal_of_minerals_metallurgy_and_materials/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:979abfac60d1ee249fff57d33c82bad27987f170a767ec033990054515d54aad
+size 63779017

data/text/iron_ore/international_journal_of_mining_science_and_technology/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:982f96560016c706f31d7cb39972c9185ae1597e75195aadfd50c718c6c59f66
+size 69378115

data/text/iron_ore/ironmaking/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8df3ccf5c44740f23370ed38e0a74ba5c9402b8488fe2bb99013d59e0ca9493f
+size 46273831

data/text/iron_ore/public_opinion/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30f91668c3cf7eefb046384f287eca10bdb2696e41a3cccaae7c233db07dee82
+size 395973820

data/text/iron_ore/sintering_and_pelletizing/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2187a656b8f7b5924d442c2d8b1c052245640f456bb9212e891b8dba6a6cdc7
+size 53992073

data/text/iron_ore/state_laws/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0be0aee4fab95a9af6c09f269866692b000bea79cd76cc407fd0606d3d19e50
+size 452632127