SI.pdf]]" - "[[raw//

- Supplementary Dataset 1.xlsx]]" ~~~ The main paper and supplements go in the same discipline folder. Supplements have no separate paper page. Omit `source_supplement` when there are none. Use this exact field name consistently. For a source-specific paper read from a full publisher or preprint webpage, record its exact URL and the reading source in `Sources`. Do not invent a local `source_pdf`. A missing PDF blocks a task explicitly based on that PDF; it does not justify pretending the source was read. For retained preprints or other versions, list their exact paths in `Sources` and include a version comparison in the body. Keep `source_pdf` as the primary paper file and `source_supplement` exclusively for supplements. ## Canonical Paper Page Use the following structure for every newly archived paper. This is a fill-in template; its bracketed prompts are instructions, not finished content. ~~~markdown --- title: "中文论文译名｜Exact English Original Title" title_zh: "中文论文译名" title_en: "Exact English Original Title" aliases: - "中文论文译名" - "Exact English Original Title" type: research tags: [academic] created: YYYY-MM-DD updated: YYYY-MM-DD sources: 1 source_pdf: "[[raw//.pdf]]" --- # 中文论文译名 **论文**：Exact English Original Title **作者**：Full author list, in source order **期刊 / 版本**：Verified journal or preprint venue and version ## 核心问题 [What question is tested, and what competing explanations matter?] ## 实验与判别方法 [Design, measurements, comparisons, sample size, and inferential logic.] ## 核心结果 [Decisive quantitative results, uncertainty, and source locations.] ## 关键解释边界 [Supported claims, assumptions, transfer limits, and unresolved inferences.] ## 后续研究 [Open questions grounded in the source.] ## 关联 [Existing wikilinks with an explanation of each connection.] ## Sources [Exact source links and the relevant pages, figures, tables, or supplement sections.] ~~~ ### Bibliographic block The field order is **论文 → DOI / 标识符, if available → 作者 → 期刊 / 版本**. - Chinese H1, original English title, full author list, and journal/version are mandatory retrieval fields. They must appear at the top of the body. - List authors in source order. Do not replace the header's list with “et al.” Affiliations may be added when useful and verified. - Journal name or preprint venue and version are the priority. Include year, volume, pages, and article number when established. Detailed submission, revision, acceptance, and online dates are optional. - Identify preprints, accepted manuscripts, reviews, and commentaries. - End each bibliographic line with two spaces to preserve separate Markdown lines. - Include an identifier only when confirmed by the source or a reliable bibliographic record. Insert it after `**论文**`, for example: ~~~markdown **DOI / 标识符**：[](https://doi.org/) ~~~ - Mirror confirmed DOI or arXiv identifiers into `doi:` or `arxiv:` when useful. - If no identifier can be confirmed, omit the entire line and property. Do not use placeholders, guessed identifiers, or incomplete DOI strings. Missing DOI alone does not make a page incomplete. **Verify article identity before copying metadata.** The top of a journal PDF may contain the end of the preceding article. Find the target paper's own title, then identify its authors and DOI in the corresponding article block. Cross-check journal, volume, page range, and supplement references. A first DOI regex match is only a candidate. **Bounded lookup for old literature.** Check the first two pages and an existing `Sources` citation once. If an identifier is absent, stop looking; do not repeatedly reopen the file or launch web searches for that missing field. If the journal cannot be recovered, state the established provenance on the journal/version line, such as `年份不详；扫描版，未见刊名`. Record the lookup outcome in the activity log to prevent repeated work. Unreadable title or authors should be reported as an identity blocker. This bounded metadata search does not shorten the full reading required for a new paper archive. ### Section requirements | Section | Required content | |---|---| | 核心问题 | The exact research question and relevant alternatives | | 实验与判别方法 | Sample and control structure, measured quantities, analytical logic, and important assumptions | | 核心结果 | Key numbers, units, uncertainty, effect sizes, or formal results, anchored to the source | | 关键解释边界 | What the design establishes, model-dependent or correlational parts, external validity, and version status | | 后续研究 | Source-grounded open questions; state plainly if the paper supplies no specific follow-up | | 关联 | Existing pages plus the specific shared mechanism, disagreement, method, or boundary | | Sources | Primary source, supplements and versions used, with precise supporting locations where possible | For reviews/commentaries, use `## 证据范围与综合方法` in the method slot. For theoretical/mathematical papers, use `## 模型、定义与证明结构` and preserve definitions, theorem conditions, quantifiers, and the proof's logical dependencies. Keep the section order. Use `## 可检验的新假说` instead of `## 后续研究` only when there is a worthwhile, evidence-grounded prediction. State the manipulation or comparison, observable result, and outcomes that would support or falsify it. Distinguish the authors' hypothesis from a wiki inference. Specialized subsections may be added. The required sections keep their names and evidentiary functions; clarifying suffixes in parentheses are acceptable. ### Evidence and interpretation - Read the complete source, including figures, captions, tables, and supplements needed for the claims being archived. Inspect rendered pages when text extraction loses structure. Disclose any unreadable material. - Preserve the numbers or formal conditions that determine the conclusion. Length alone does not establish depth. - Separate author findings, author interpretation, and wiki synthesis. - In `关键解释边界`, describe the actual limitation and its consequence. Do not add severity scores, repeated emphasis, or invented flaws. - Attribute self-declared limitations to the authors. - Derive or calculate the direction of an alleged bias before asserting which conclusion it favors. - Keep observation, proxy, model estimate, causal claim, and speculation distinguishable. - Create concepts, topics, and hypotheses only when they add a supported distinction or connection. - If no useful connection exists yet, say so and link the paper through the index. Do not invent a related page. ### Math and layout Use `$...$` for inline math and `$$...$$` for display math in Obsidian. Do not use `$...$` or `\[...\]`. Check paired delimiters, subscripts, superscripts, braces, units, and source fidelity. Use short numbered steps or a table for processes when they read more clearly. For aligned mathematical expressions, use an `aligned` environment inside a display block. Avoid long ASCII diagrams that rely on spaces and font width. View important formulae and diagrams in Reading view; raw Markdown alone cannot verify their appearance. ## Inbox Workflow ### 1. Inventory and read List every incoming file, including PDF, DOCX, spreadsheet, and other supplements. Identify main paper, supplement, potential duplicate, and version relationships. Read each source fully for a new archive. Match incoming papers against the index and relevant existing pages. Record the question, contribution, key quantitative evidence, useful connections, supported disagreements, and proposed discipline folder. If a supplement arrives later, locate its parent paper and update that page. Reconsider claims affected by the added evidence. ### 2. Run all five duplicate checks Run the following before creating a `research-` or `resource-` page. Repeat the identifier audit across the live wiki after the batch. | Check | Required action | What it detects | |---|---|---| | Title and author | Search the index and live pages for title keywords, author/year, and the surname alone | Existing summaries with different naming or damaged title text | | Source path | Search `source_pdf` and other source fields for the exact incoming/final path | A source already registered | | Slug prefix | Search `research---` in existing filenames | A paper filed under another keyword suffix | | Content hash | Compare the incoming file's MD5 against files of the same kind throughout `raw/`, excluding its own path | Byte-identical files with different names | | DOI / persistent identifier | Compare confirmed identifiers, then audit all live paper headers after the batch | Publisher variants with different bytes and the same identity | MD5 is used for accidental duplicate detection. A hash mismatch does not establish a different paper. Author/year matches and shared DOI values are candidates that require source verification; a stored identifier may itself be wrong. Compare preprint and formal versions even when titles match or identifiers differ. Record changes to methods, samples, figures, claims, discussion, and limitations. Preserve useful removed arguments as version-specific statements. Do not infer why a change occurred unless the authors or review record explain it. If the same source already has a page, update that page. Keep incoming duplicate files untouched when the user asks to retain them. Never delete a duplicate on the agent's initiative. ### 3. Present the batch plan List each source, destination, final filename, page to create/update, relevant topic/toolbox changes, and exceptions. State the total file scope. Ask for confirmation for three or more affected files, bulk moves, or directory restructuring. Ask for a new discipline folder name when needed. Continue work already approved for existing folders while a separate folder decision is pending. ### 4. File sources and write pages After confirmation: 1. Find every existing wikilink to a file scheduled for moving or renaming. Prepare its path replacements before moving. 2. Move accepted papers from `raw/inbox/` into the approved discipline folder and apply all affected link updates as part of the same operation. 3. If renaming was approved, use the original paper title as the basis. Apply only necessary filename-safe substitutions and practical shortening; preserve the exact title in the note. Never overwrite an existing destination. 4. Move supplements with the paper. Give them identifiable names such as `

SI.pdf` or `

- Supplementary Dataset 1.xlsx`. 5. Check the final paths, then create or update the canonical paper page. 6. Update relevant topic/concept pages only where evidence warrants it. 7. Update the academic toolbox when there is a substantive methodological contribution. 8. Update `index.md` and `overview.md`, then append to `log.md`. Recompute counts. If any operation fails, record the actual state and repair affected references before reporting success. Never claim a file was filed merely because its intended path was written into a note. ### 5. Audit and report Run the completion checks below. Report files processed, actual destinations, pages created/updated, version comparisons, unresolved issues, and retained files. Inbox completion means every file has an explicit disposition. Unreadable sources, unresolved identities, unapproved destinations, and files the user asked to retain may remain in inbox. List these exceptions. Do not move an unresolved file solely to empty the folder. ## Academic Toolbox Maintain `wiki/topic-academic-methodology.md` as a lookup from research question to method and evidence. ~~~markdown ### | 我要回答的问题 | 候选方法及检测对象 | 适用条件与解释边界 | 来源 | |---|---|---|---| | | | | | ~~~ A method row should explain what is measured, which inference it supports, and where it fails to identify the intended target. Add rows for substantive new methods, validation, or changed interpretation. A paper's routine use of a method alone need not create another row. Reuse the relevant section and existing row where appropriate. Add a domain section only when needed. Keep source counts tied to distinct papers, not the number of rows or edits. ## Other Workflows ### Ingest For a provided full paper, file, webpage, pasted text, or image: read the available source, determine its type and scope, perform the relevant duplicate checks, and apply the same approval, evidence, registration, and audit rules. Limited supplied content supports a limited resource note; do not claim full-paper coverage. For every completed ingest, update the index and overview and append a log entry. Add methodological evidence to the toolbox where relevant. ### Query Start with `index.md`, then read relevant pages and, where needed, their original sources. Answer with source links. Save a reusable synthesis as a concept/topic/resource page when requested or included in the approved scope, and register it in the index and log. ### Bibliography-only repair When asked only to make older pages searchable, restore the Chinese/English title, authors, and journal/version. Apply the bounded identifier lookup. Preserve existing analysis. Do not expand the task into full rereading or mass rewriting. ### Lint Inspect broken links, missing source paths, orphan pages, index omissions, duplicated identities, YAML, section structure, stale claims, and count drift. Report findings with file locations. Apply obvious repairs within authorized scope. Discuss changes that alter interpretation; do not fabricate content to satisfy a checker. Legacy pages are not silently rewritten to meet a new template. Bring them into the current format when substantively edited or explicitly included in a migration. ## Completion Checks Checks must cover both metadata and content. Report unreadable files and exceptions; a checker that skips them cannot establish a clean result. | Check | Passing condition | |---|---| | YAML | Frontmatter parses; dates, lists, titles, and source fields have the intended types | | Bilingual identity | Chinese H1, `title_zh`, `title_en`, bilingual `title`, and both aliases agree | | Bibliography | Original title, full author list, journal/version, correct order and line breaks | | Identifier | Every included identifier belongs to this paper; absence alone is accepted | | Source PDF | Every local PDF-derived page has an exact, existing `source_pdf` | | Supplements | Every listed path exists; every supplement filed in this batch is indexed on its parent page | | Structure | Required sections appear in order, accepting the documented method and follow-up alternatives | | Evidence | Key numbers, units, uncertainty, conditions, and claims match the source | | Version | Preprint/formal status is explicit; available differences are recorded | | Links | Inbound replacements and outbound links resolve; every new page is registered | | Deduplication | All five checks were completed and identifier collisions investigated | | Bookkeeping | Index, overview, toolbox where applicable, and append-only log reflect the batch | | Counts | `total_pages` is recomputed from live files, excluding the `_archive` directory | | Presentation | Important math, tables, and diagrams have been visually inspected where possible | Use a YAML parser for frontmatter validation. Read an entire quoted `source_pdf` value before removing its outer wikilink delimiters; filenames can contain `]`, as in isotope labels. Do not use a parser that stops at the first closing bracket. Ignore documentation examples and code fences when auditing live wikilinks. Resolve vault-relative paths, optional aliases, headings, and file extensions. Check actual paths rather than accepting any filename with a matching basename. ## Index, Counts, and Log Keep a catalog row for each live content page: ~~~markdown | 页面 | 类型 | 内容 | |---|---|---| | | research | | ~~~ `total_pages` in `wiki/index.md` equals the number of live `wiki/**/*.md` files, including index, log, overview, and toolbox. Exclude files whose path contains an actual `_archive` directory component. Recompute after each batch: ~~~bash python -c "from pathlib import Path; print(sum(1 for p in Path('wiki').rglob('*.md') if '_archive' not in p.relative_to('wiki').parts))" ~~~ Do not hand-increment this value. File count, catalog rows, and source count measure different things. Append log entries using this shape: ~~~markdown ## [YYYY-MM-DD] ingest | Pages created: X | Pages updated: Y Sources filed: X | Files retained in inbox: Y Key additions: Checks: Remaining: ~~~ Record old-literature metadata searches once so future passes do not repeat them. Keep any historical log entry intact. ## FATAL Rules These constraints govern the workflows in this file. - **FATAL-001: Plan bulk changes.** State the concrete plan and obtain confirmation before bulk moves or directory restructuring. - **FATAL-002: Preserve sources and knowledge.** Never modify or delete source-file contents. Never delete wiki content; archive it with links preserved. The normal source-file operation is an approved move from inbox to a discipline folder, including a reviewed filename change and matched supplements. Any later reorganization of filed sources requires explicit authorization. - **FATAL-003: Preserve references.** Locate every inbound wikilink before a rename or move, update affected links with the change, and verify resolution afterward. - **FATAL-004: Confirm batches.** Changes affecting three or more files follow PLAN → CONFIRM → EXECUTE. Do not ask again for steps already covered by the user's confirmed scope. - **FATAL-005: Preserve source traceability.** PDF-derived paper pages need an exact, existing `source_pdf`. Index every filed supplement. Never manufacture source paths. - **FATAL-006: Check identity five ways.** Check title/author, source path, slug prefix, content hash, and persistent identifier. Update an existing page for the same source. Compare versions before consolidating their evidence. - **FATAL-007: Keep academic evidence distinct.** Social posts, personal accounts, promotional commentary, and social-domain wiki pages cannot substantiate academic claims. Trace such claims to primary research or leave them out of the academic record. Describe mechanisms and evidence in plain language; avoid rhetorical reversals, slogans, and manufactured criticism. - **FATAL-008: Preserve the paper template and evidence density.** Keep retrieval fields, section order, source-grounded detail, quantitative conditions, interpretation boundaries, and accurate version labels. Do not archive a full available paper from its abstract alone. Do not promote a plausible mechanism to a reported result. Do not force a new concept, hypothesis, disagreement, or link merely to populate a section. ## Purpose Template Create `purpose.md` from this section during confirmed initialization. Replace prompts with the user's actual scope. ~~~markdown # 知识库用途 ## 研究方向 [Describe the research questions this vault should help answer.] ## 收录范围 [Paper types, disciplines, languages, and other permitted primary sources.] ## 使用方式 [For example: literature review, comparing methods, tracing evidence, developing research questions.] ## 内容标准论文正文使用中文，保留英文原题。事实、作者解释和跨论文推断分别标明。每项重要判断可以追溯到原始来源。每轮归档更新目录、总览和活动日志。 ## 已确认的学科目录 | Folder | Scope | |---|---| | | | ## 当前偏好与例外 [For example: duplicate files to retain in inbox; approved naming rules; current task limits.] ~~~ Update purpose and approved exceptions when the user changes their preferences. Keep the research questions and editorial standards specific enough to guide the next session.