Data Verification & Provenance

1. Overview · 信任基石：以透明替代权威

SemiconFab operates on a foundational philosophy: trust in industrial data is earned through complete methodology transparency rather than centralized claims of institutional authority. In semiconductor manufacturing, where access to definitive equipment datasheets is restricted by commercial practices and confidential agreements, asserting that an entire database is infallible is both unrealistic and counterproductive. Instead, SemiconFab isolates and tags every single specification parameter with its exact evidentiary origin, verification timestamp, and confidence tier. By openly distinguishing between vendor-published figures, engineering extrapolations, and simulation placeholders, we provide cleanroom planners and researchers with the unvarnished context required to make informed engineering judgments.

SemiconFab 坚守的核心理念是：工业数据的信任源自方法学的全面透明，而非中心化的机构权威背书。在半导体制造领域，由于商业保密协议与市场壁垒，获取完全权威的原厂设备技术白皮书极其困难，宣称某套数据库“100% 绝对权威”既不现实，也不负责。SemiconFab 的做法是：对每一项规格参数进行颗粒度解耦，为其标注独立且明确的数据来源出处、核验时间与置信等级。通过将原厂公开指标、工程推断值与模拟占位符公开区分，我们为晶圆厂规划工程师与科研人员提供真实无修饰的工程背景，支持其做出严谨的工程判断。

2. The 5 Provenance Tiers in Detail · 五级出处标准详析

Every quantitative and qualitative value in the catalog carries an immutable provenance.tier attribute. The table below details the evidentiary criteria, assignment authorities, and representative examples for each tier:

目录中的每一个具体参数数值均携带不可拆分的 provenance.tier 出处属性。下表详述各层级的判定证据标准、赋权责任主体与典型工程范例：

3. Submit-to-Publish Workflow · 从提交到发布的五步流水线

To prevent unvetted commercial claims from polluting the catalog while maintaining low submission barriers, records pass through an automated five-stage pipeline:

为杜绝未经证实的商业宣传污染公共数据库，同时保持开放的提报门槛，所有设备规格记录均须经过严格的五阶段流水线：

[1. Form Submission] ──> [2. Schema Validation] ──> [3. Tier Gating] ──> [4. Human Audit] ──> [5. Publication]
    网页表单提报              格式与范围自动化校验          默认赋予 inferred       原厂声明人工核验         带标公示发布

• Step 1: Submission via Web Form (网页表单提报): Vendors, facility engineers, and researchers enter equipment specifications into the structured web submission form (/add-equipment.html). The form dynamically loads schema.json rules, ensuring data is captured in canonical SI units with mandatory field-level provenance metadata.
  设备厂商、厂务工程师或研究人员在结构化网页表单（/add-equipment.html）中填写设备参数。表单动态加载 schema.json 规则，强制要求所有数值采用法定 SI 单位制并附带字段级出处元数据。
• Step 2: Automatic Schema Validation (自动化模式校验): Upon submission, an automated validation worker parses the payload against schema.json. It verifies JSON syntax, validates regex patterns (e.g. UUID v4 identifiers), checks mandatory layer groupings for the declared conformance level (L1/L2/L3), and confirms SI unit keys. Malformed records are immediately rejected with descriptive field paths.
  后端校验工作流使用 schema.json 对提交数据包进行语法解析。严格核验 JSON 格式、正则表达式（如 UUID v4 记录号）、所选合规分级（L1/L2/L3）对应的必填数据层完备性及单位合法性。不合规记录即刻拦截并返回精确定位到字段的报错信息。
• Step 3: Tier Assignment & Anti-Spam Gating (出处赋权与防刷关机制): Every unverified field is automatically assigned the inferred tier by default. The pipeline never automatically grants vendor_confirmed or independent_verified status. Even if a submitter checks "Official Vendor", claims of primary authority cannot bypass human scrutiny.
  所有未经核验的字段在入库时一律默认赋予 inferred 等级。系统流水线绝不允许自动化授予 vendor_confirmed 或 independent_verified 标识。 即使提交者自行勾选“厂商官方规格”，任何关于原厂官方背书的声明都绝不可能绕过人工审查自动生效。
• Step 4: Human Review of Official Claims (原厂资质人工抽检): Records claiming vendor_confirmed status are placed in an audit queue. A catalog maintainer checks that the submission originates from an official corporate domain email, confirms that provided source URLs resolve to public technical manuals, and spot-checks key physical dimensions against public documentation. Submissions without verifiable corporate credentials remain published as inferred.
  声明 vendor_confirmed 等级的记录进入待审队列。维护人员通过两项标准进行核查：提交人必须使用企业实名域名邮箱、提供的来源链接必须直达公开技术手册，并对机台外形几何尺寸进行抽检比对。无法出具企业原厂有效证明的提报将保持为 inferred 发布。
• Step 5: Catalog Publication with Tier Badges (带有出处徽章正式发布): Approved records transition to published status and are merged into the public catalog. Every displayed field features an explicit provenance dot, allowing readers to instantly discern verified specifications from estimated values.
  审核通过的记录状态变更为 published，正式并入公共目录库。网页前端展示的每个技术指标旁均附带高对比度的出处状态徽章，读者可瞬间识别各项指标是原厂确认还是工程估算。

Figure 2: The end-to-end SemiconFab submission and verification pipeline, illustrating automated schema parsing, strict tier assignment gating, and the ongoing community correction loop.

图 2：SemiconFab 设备数据提交与核验完整流水线，展示自动化模式校验、严格出处赋权防刷关机制及上线后的社区纠错闭环。

4. Tier Upgrade Path · 出处升级机制

A record in the catalog is not static; it improves over time as cleanroom operations generate additional empirical data. To ensure this process can be maintained without administrative overhead, upgrade criteria are designed to be concrete, deterministic, and lightweight:

目录中的设备记录并非静态快照，而是随着产线运行数据的积累持续演进。为确保单人维护者亦能高效运作该机制，升级规则设计得直观、确定且免除官僚冗余：

• Upgrading from inferred to community_corroborated (升至社区交叉确认):

  Requirement: Two independent engineering submissions from distinct organizations matching within a 5% measurement tolerance (for numerical values) or identical configurations (for categorical options).

  来自两个不同法人主体的工程师独立提报，数值偏差在 5% 以内或枚举配置完全一致。当第二位独立技术人员提供相互印证的工程数据时，维护人员确认无关联后即可升档为 community_corroborated。

• Upgrading from inferred` to `vendor_confirmed (升至厂商公开确认):

  Requirement: Official confirmation by the equipment manufacturer via an official datasheet link, product whitepaper, or corporate email domain attestation.

  设备制造商官方正式背书。厂商认证代表提供官方产品白皮书公开链接或通过企业实名邮箱出具规格说明单，维护人员存档链接后将相关字段晋升为 vendor_confirmed。

• Upgrading to independent_verified (升至独立第三方验证):

  Requirement: A public, citable measurement report from an ISO-accredited testing body, cleanroom facility acceptance certificate, or published academic literature.

  提供具备国家或国际认证资质的检测机构检测单、晶圆厂装机验收签字文件或学术论文 DOI 编号。核查真实有效后赋予 independent_verified。

5. Catalog Badges & UI Indicators · 页面徽章与数据完整度标识

When browsing the equipment catalog, readers are presented with clear visual indicators reflecting record maturity and field accuracy:

用户在浏览设备详情页时，系统通过标准化的视觉徽章实时标识整机完备度与字段级可信度：

• Record-Level Conformance Badges: L1: Minimum Listing (green), L2: Fab Planning (teal), L3: Benchmark Grade (indigo), and Draft (gray).
• Field-Level Provenance Dots: 🟢 Green (vendor_confirmed), 🔵 Blue (independent_verified), 🟡 Yellow (community_corroborated), 🟠 Orange (inferred), ⚪ Gray (synthetic).
• "Help Complete This Record" Call to Action: On records with incomplete layers (such as the 114 draft tools awaiting footprint), interactive prompts highlight missing parameters to encourage community contributions.

6. Public Corrections & Dispute Resolution · 公开纠错与争议处理

Transparency requires a rapid, accessible mechanism to dispute and correct erroneous data. Anyone—including competing vendors, fab operators, or researchers—can challenge an inaccurate value in the catalog:

真正的透明度必须包含便捷的公开纠错与争议复核机制。无论是设备厂商、晶圆厂运行一线工程师还是行业研究员，只要发现目录中的数据有误，均可随时提起纠错争议：

• How to Submit a Correction (如何提交纠错): Click the "Flag Error" button located beside any parameter on an equipment detail page, or email our maintainers at [email protected]. Please include the record_id, the specific field key, your proposed correction, and supporting documentation (e.g., official datasheet PDF, field photograph, or measurement log).
  点击设备详情页中任意字段旁边的“报错纠偏”按钮，或直接发信至平台维护邮箱 [email protected]。邮件请附上 record_id、争议字段名、建议更正值以及支撑证据材料（如公开手册 PDF、工程实拍图或测试记录）。
• Dispute Gating Policy (争议处置原则): Upon receipt of a substantiated correction claim, the disputed parameter is immediately downgraded to inferred with an annotation noting ongoing review. If the manufacturer or submitter fails to substantiate the original figure within 14 calendar days, the catalog adopts the corrected value backed by the new evidence. All revision histories are recorded immutably in the record changelog.
  收到具备支撑证据的纠错请求后，争议数值将立即降级标记为 inferred 并标注复核中。若原数据提报方在 14 个自然日内无法提供更权威的反驳凭证，系统将依据新证据正式采纳更正数值。所有的更正历史均永久记录于该条记录的版本变更日志中。