---
title: "Journal Platform Research"
canonical_url: "https://scholarxiv.com/developers/docs/journal-platform-research"
markdown_url: "https://scholarxiv.com/developers/docs/journal-platform-research.md"
---

> For the complete documentation index, see [llms.txt](/llms.txt).

# Journal Platform Research
URL: /developers/docs/journal-platform-research
LLM index: /llms.txt

# Building a Journal Inside ScholarXIV — Research Dossier

Everything a submission → vetting → publication pipeline actually consists of, reverse-engineered
from the real systems: **arXiv**, **Nature (Springer Nature)**, **bioRxiv / medRxiv**, **PLOS ONE**,
**eLife**, **F1000Research**, and the three manuscript systems that run most of the industry
(**ScholarOne Manuscripts**, **Editorial Manager**, **Open Journal Systems**).

Sources are primary — publisher/server author guides fetched July 2026. Where a number is quoted
(word counts, dpi, character limits) it is quoted from the source, not from memory.

---

## 0. TL;DR — the three archetypes

Every venue in scholarly publishing is one of three machines. Pick which one(s) we are building
before anything else, because the data model and the UI diverge hard after the submit button.

| | **A. Preprint server** | **B. Selective journal** | **C. Publish-then-review** |
|---|---|---|---|
| Examples | arXiv, bioRxiv, medRxiv | Nature, PLOS ONE, Science | F1000Research, eLife (new model) |
| Gate before public | Moderation/screening (hours–days) | Peer review (months) | Prepublication checks (days) |
| Reviewers | none | 2–3 invited experts | invited, **after** publication, reports public |
| Outcome | posted / declined | accept / revise / reject | always published; review status is a badge |
| Versions | v1, v2, v3… same DOI | R1, R2… one final VOR | Version 1, 2, 3 each with own DOI |
| Cost to author | free | $0–$11.5k APC | APC |
| Rejection exists? | only for scope/ethics | yes, ~92% at Nature | no — only "not indexed" |

**My recommendation for ScholarXIV:** build **A** first (it is a 2–3 week feature and it works with
the LaTeX workspace you already have), design the schema so **C** is a config flag on top of it, and
treat **B** as the paid/institutional tier later. Rationale in §19.

---

## 1. The universal object model

Strip away branding and every one of these systems has the same seven objects. Build these and you
can express any of the three archetypes.

```
User ──┬── Author (submitting / corresponding / co-author)
       ├── Reviewer
       ├── Editor (Editor-in-Chief / Handling / Section / Guest)
       ├── Moderator (preprint-only role)
       └── Production staff (copyeditor, layout/typesetter, proofreader)

Submission ──1:N── Version (v1, v2 / R1, R2)
Version    ──1:N── File (manuscript, figure, table, supplementary, source, cover letter…)
Version    ──1:1── Metadata record (the "form")
Version    ──1:N── ReviewRound
ReviewRound──1:N── ReviewAssignment ──0:1── ReviewReport
ReviewRound──1:1── Decision
Submission ──0:1── PublicationRecord (DOI, license, issue, galleys, indexing)
Submission ──1:N── Discussion thread (author↔editor, editor↔editor, editor↔reviewer)
Submission ──1:N── AuditEvent (the activity log — legally and ethically load-bearing)
```

Two things people always get wrong on the first build:

1. **Metadata is versioned, not attached to the submission.** Titles, author order, and abstracts
   change between rounds. Nature explicitly rewrites summary paragraphs after acceptance.
2. **Files have a *designation*, not just a name.** ScholarOne calls it "file designation",
   Editorial Manager calls it "item type". This is what decides: does it go in the reviewer PDF?
   Is it anonymized? Does it go to production? See §12.4.

---

## 2. arXiv — complete walkthrough

The most implementable model, and the one your users already live in.

### 2.1 Prerequisites (before you can even start)

| Gate | Rule |
|---|---|
| Account | Must be a registered arXiv user. |
| **Endorsement** | First-time submitters, and anyone posting to a **new category**, must be endorsed. |
| Automatic endorsement | Granted if you have claimed ownership of papers submitted by co-authors **and** have an institutional email address. |
| Manual endorsement | Start a submission → pick category → arXiv emails you an endorsement request with a shareable code (6-char alphanumeric). You find an eligible endorser (abstract pages carry a "Which authors of this paper are endorsers?" link), they submit **positive / negative / abstain**. Need ≥1 positive **per category**. |
| Endorser eligibility | Must have authored enough papers in that "endorsement domain"; **only papers submitted 3–5 years ago count**. Calibrated so any active scientist of a few years can endorse. |
| Explicitly *not* | "Endorsement is not peer review." Endorsers verify the person is a real researcher in that field — not that the paper is correct. |

### 2.2 The submission wizard, screen by screen

1. **Start** — "START NEW SUBMISSION" on the user page.
2. **Prepare Files** — choose file(s); zip/tar.gz for multi-file. Upload. Delete extraneous files
   (trashcan icon / "Delete All").
3. **Check Files** — automated analysis detects: the correct processor/compiler, the top-level TeX
   file, and extraneous files it recommends deleting.
4. **Compiler & top-level file** — auto-detected compiler shown in a dropdown, overridable. Only
   `.tex` files containing `\documentclass` are offered as top-level. Multiple top-level files can
   be selected → concatenated output.
5. **Auto-detected Notes** — review recommended deletions; uncheck what you want to keep.
   "Return to Upload Step" available.
6. **Accept and Continue** → confirmation popup for deletions.
7. **Compile & preview** — success/failure with the full log. On success, preview the PDF.
8. **Metadata** (§2.3).
9. **License** (§2.4).
10. **Agreements** — Submittal Agreement, Code of Conduct, Moderation policies, Privacy policy.
11. **Review & Submit Article** — the timestamp is set here.
12. **Unsubmit** is available up to announcement (returns it to incomplete; no new version is
    generated for pre-announcement edits).

### 2.3 arXiv metadata fields — exact spec

All fields are **ASCII only**; Unicode must be converted to LaTeX. ("Bad character" errors almost
always come from pasting curly quotes/em-dashes/ligatures out of a PDF.)

| Field | Req | Rules |
|---|---|---|
| **Title** | ✔ | No ALL CAPS, no Unicode. Limited MathJax + LaTeX accent commands. Expand cryptic macros. May reference `arXiv:YYMM.NNNNN`. |
| **Authors** | ✔ | `Firstname Lastname` or `Firstname Middlename Lastname`. No honorifics. Initials with periods allowed. No ALL CAPS. No commas or the word "and" inside a name. Suffixes attach without comma (`John Smith Jr`). Affiliations in parentheses, **city/country only** — no street/zip. Numbered form: `Author One (1), Author Two (1 and 2) ((1) Inst One, (2) Inst Two)`. Collaboration names allowed. |
| **Abstract** | ✔ | **Max 1920 characters.** Must not contain the word "Abstract". Some LaTeX via MathJax. Carriage returns stripped unless followed by leading whitespace. Strip `~`, `\,`, `\ `, `\em`, `\it`, font commands. |
| **Comments** | ○ | Page + figure count; "to be published in…", "submitted to…", conference info; author roles ("Appendix by…"). For replacements, describe the change. URLs need a trailing space. **No copyright statements.** |
| **Categories** | ✔ | One **primary** (drives the announcement mailing) + optional **cross-lists**. |
| **Report-no** | ○ | Institution-assigned number only, e.g. `EFI-94-11`. Nothing else. |
| **Journal-ref** | ○ | Only if already published. Volume, year, pages. Multiple separated by `; `. **No URLs.** |
| **DOI** | ○ | DOI of *another version* of the article. Do **not** put the arXiv-assigned DOI here. Space-separated for multiples. |
| **MSC-class** | ○ | Math only. `14J60 (Primary) 14F05, 14J26 (Secondary)`. |
| **ACM-class** | ○ | CS only. `F.2.2; I.2.7`. |
| **"Are you an author?"** | ✔ | Yes/no. Submitter becomes the article *owner* with edit rights regardless. |

### 2.4 arXiv licenses (irrevocable — cannot be changed later)

- CC BY 4.0
- CC BY-SA 4.0
- CC BY-NC-SA 4.0
- CC BY-NC-ND 4.0
- CC0 1.0 (public domain)
- arXiv.org Perpetual, Non-Exclusive License 1.0 (minimum grant: arXiv may distribute, nothing more)

All **metadata** is CC0 regardless of the article license.

### 2.5 File formats

| Preference | Format |
|---|---|
| 1 | (La)TeX, AMS(La)TeX, PDFLaTeX |
| 2 | PDF (author-authored, not TeX-derived) |
| 3 | HTML with JPEG/PNG/GIF |
| **Rejected** | DVI; PS; **PDF generated from TeX/LaTeX source**; scanned documents in any format |

- Figures: **PS/EPS** for the LaTeX path; **JPEG/GIF/PNG/PDF** for the PDFLaTeX path. Mixing the two
  is the #1 compile failure.
- Filenames: only `a-z A-Z 0-9 _ + - . , =`. **Case-sensitive** — `Figure1.PDF` ≠ `figure1.pdf`
  (the #2 compile failure).
- **Submissions with omitted figures are not accepted**, even with external links.
- Size: no published hard cap; from Feb 2026 arXiv warns on images >34 megapixels (≈ A4 @ 600dpi).
  Oversized submissions get an automatic rejection identifier; you email admins with that ID to
  request an exception. Guidance: JPEG for photos, PDF/PNG/GIF/PS for line art.

Common failure list arXiv itself publishes: mixed figure formats; filename case mismatch; missing
custom `.sty`; missing/misnamed/wrong-path files.

### 2.6 Moderation (the vetting model to copy)

- **Who:** volunteer subject-matter experts with terminal degrees, approved by discipline advisory
  committees + arXiv staff. Confidential, treated like journal peer review.
- **Criteria:** (1) *scholarly standards* — real sections, figures, references, careful prep;
  (2) *scholarly interest* — originality/novelty/significance; no falsified or plagiarized material;
  (3) *content type & format* — research articles primary; no egregiously offensive images;
  **significant generative-AI use must be disclosed**; (4) *scope* — must fit the taxonomy and serve
  a represented community; (5) *originality* — no duplicated content; **max 3 back-catalogue papers
  per day** per submitter.
- **Outcomes:** Accept · Reclassify (move/cross-list) · Decline (pre-announcement) · Withdraw
  (post-announcement policy violation) · **Hold** (extra scrutiny for authors with prior rejections).
- **Appeals:** exist for classification and decline; **decisions on appeal are final, no feedback
  given**; some appeals require prior journal publication.
- **Timing:** submitted by 14:00 ET → generally available 20:00 ET.

### 2.7 Lifecycle after announcement

- **Replace** (not withdraw) for updates → new version v2, v3…; all prior versions stay publicly
  downloadable forever.
- **Withdraw** creates a new version *marked withdrawn*, displaying the reason and not linking to
  full text. Valid reasons: fundamental error, legitimate scientific concern. **Invalid** reasons:
  "it got published" (use journal-ref), "I updated it" (use replace), copyright (separate process).
- Nothing is ever truly deleted post-announcement.

---

## 3. Nature — complete walkthrough

The reference implementation for a highly selective journal. **~8% acceptance; most submissions are
declined without external review.**

### 3.1 Initial submission — deliberately loose

Nature is "flexible with regard to the format of initial submissions… style and length will not
influence consideration." What they actually want:

| Item | Spec |
|---|---|
| Container | **One single PDF or Word file, ≤30 MB**, text + figures together, figure legends on the same page as their figure. |
| Line numbers | Required. Word docs are auto-numbered by the system; if you supply PDF **you** must number the lines. |
| LaTeX | **Not accepted at initial submission** — PDF only. LaTeX is accepted at acceptance stage. |
| Title | ≤ **75 characters incl. spaces** (two print lines). No technical terms, abbreviations, or active verbs. |
| Authors | Corresponding author(s) marked with `*`. **LLMs cannot be authors.** LLM use documented in Methods. |
| Abstract / summary para | Fully referenced, ideally ≤ **200 words**, structured as: 2–3 sentences of basic-level intro → background & rationale → main conclusions ("Here we show…") → 2–3 sentences of general context. |
| Main text | ~**2,500 words + 4 display items** (typical 6-page article) or ~**4,300 words + 5–6 display items** (8-page). A "modest" display item ≈ ¼ page ≈ 270 words; a composite figure ≈ ½ page ≈ 600 words. Subheadings ≤ 40 chars. |
| References | ≤ **50** in main text; article titles required; one publication per number; superscript numerals; **no EndNote linked fields**; all authors listed unless >5 → first author + et al. Preprints and DOI'd datasets are citable. |
| Methods | Typically ≤ **3,000 words**; Methods-only refs don't count; no figures/tables allowed inside Methods. |
| Extended Data | Up to **10** multi-panel items — online only, not in print. |
| Supplementary Info | Only material too large/impractical for print; **should not contain figures** (those become Extended Data). |
| Cover letter | **Optional.** Not shared with referees. Use it for: why it matters, conflicts of interest, related work in press/submitted elsewhere. |
| Related manuscripts | **Mandatory disclosure** of any overlapping-authorship manuscript under consideration (incl. under appeal) or in press elsewhere — copies uploaded as separate marked files. |
| Double-anonymized | Opt-in checkbox at submission; **anonymization is the author's responsibility**, editors don't verify. |

### 3.2 Required statements (initial submission)

- **Author contributions** — per-author, by initials. Mandatory.
- **Competing interests** — mandatory for *all* content.
- **Funding** — separate statement, only if the work arose directly from a specific grant.
- **Data Availability** — in Methods, under that exact heading. Certain data types *must* be
  deposited in a structured public repository with accession numbers, fully open at publication.
- **Code Availability** — required whenever custom code is central to the conclusions; editors may
  decline the paper if important code is unavailable.
- **Materials & Correspondence** — who handles requests.
- **Acknowledgements** — brief, no thanks to referees/editors, no effusive adjectives.
- **Reporting Summary** — life sciences + behavioural/social sciences: a completed PDF form,
  available to editors and reviewers during assessment, **published with the accepted paper**.
  (Uses advanced Acrobat features — must be filled in Adobe Reader.)
- **Discipline checklists**: CONSORT (phase II/III RCTs — non-conforming reports may be sent back
  before review), REMARK (tumour-marker prognostic studies), BRISQ Tier 1 (human biospecimens),
  trial registration in ClinicalTrials.gov or an ICMJE-recognized registry **before enrollment**.
- **Crystallography**: `.cif` file + structure factors (embedded, or `.hkl`/`.fcf`) + CheckCIF PDF
  output explaining every A/B-level alert + probability-ellipsoid figure in the SI.
- **Compound numbering**: all inorganic/organic compounds bold-numbered in order of appearance.

### 3.3 Statistics reporting (they are strict)

Methods must carry a statistics subsection stating each test and whether one- or two-tailed; error
bars defined in every figure legend; **exact n** for every statistic (individual values, not a
range); number of repeats for representative results; exact P values for significant *and*
non-significant results; F and df for ANOVA; t and df for t-tests; replicates explicitly defined.

### 3.4 The editorial process

```
Submit
  └─ Editorial assessment (subject editor + informal advice from scientific advisors/colleagues)
        ├─ Decline without review  ← the majority. Author usually informed within ~1 week.
        └─ Send to peer review (2–3 referees typical; more for stats/technique specialists)
              ├─ Referee reports (~2 weeks agreed turnaround, submitted online)
              └─ Editor + editorial colleagues confer → decision letter, one of five outcomes
```

**Criteria for publication:** original research (not published/submitted elsewhere) · outstanding
scientific importance · conclusion of interest to an interdisciplinary readership. Special cases:
large-dataset papers need comprehensiveness or a technical/scientific advance; purely technical
papers need broad community impact; therapeutic papers without mechanism need major disease impact.

**Referee selection criteria (verbatim):** independence from authors and their institutions ·
ability to evaluate technical aspects fully and fairly · currently/recently assessing related
submissions · availability within the requested time. Authors may suggest referees (with contact
details) and may **exclude one or two (no more) competing groups**; editors decide. Potential
referees must disclose professional/commercial competing interests before agreeing, and must not
circulate the manuscript.

**The five decision outcomes:**
1. Accept, no further changes.
2. Accept in principle after revisions — usually **not** re-sent to referees; editor checks.
3. Decision deferred pending revision — further experiments needed; **is** re-sent to some/all
   referees; requires a point-by-point response to every comment.
4. Reject, but resubmission would be considered — letter states this explicitly; resubmission must
   address *all* technical comments + point-by-point response, else it isn't re-reviewed.
5. Reject outright, no resubmission — appeal only in writing, on scientific grounds; **manuscript
   may not be submitted elsewhere while an appeal is pending**; new referees are rarely brought in;
   decisions are weighted by referee expertise, "not by a voting procedure."

**Reviewer questions Nature actually asks** (this is your reviewer form):
- *Key results* — outstanding features of the work.
- *Validity* — flaws that should prohibit publication.
- *Originality and significance* — with references if not original; interest to own discipline /
  multiple disciplines.
- *Data & methodology* — validity of approach, quality of data and presentation; **reviewers are
  expected to review Extended Data and Supplementary Information too**; is it detailed and
  transparent enough to reproduce?
- *Appropriate use of statistics and treatment of uncertainties* — error bars defined; comment on
  the appropriateness of each statistical test.
- (plus) confidential comments to the editor — welcome, but **must not contradict** the comments to
  the authors.

**Transparent peer review:** for manuscripts submitted since Feb 2020, reviewer comments + author
rebuttals of published research articles are published as a supplementary peer-review file (opt-in;
research Articles only; not for transferred versions reviewed at another Nature journal).
**Reviewer recognition:** every peer-reviewed paper carries an acknowledgement statement; reviewers
may consent to being named ("Nature thanks X, Y and Z for their contribution to the peer review of
this work" / "…and the other, anonymous, reviewer(s)…").

**Transfers:** a decline letter carries a link to transfer all materials + decision letter +
referee comments to another Springer Nature journal without re-entering submission info. Referee
comments travel; **between journals with in-house editors, referee identities travel too**.

### 3.5 Final (post-acceptance) submission — this is where the real forms appear

Text and display items now go up as **separate files**, plus scanned signed forms.

- **Text:** Microsoft Word preferred, style tags removed, 12pt Times New Roman, single carriage
  returns, no indents/page breaks, real super/subscript formatting, Symbol font or normal text for
  Greek/math, Equation Editor/MathType only where unavoidable.
- **LaTeX:** accepted now, but **converted to Word before typesetting**. Rules: any standard class
  (`article.cls`, `revtex.cls`, `aastex.cls`, `amsart.cls`); **remove all personal macros**; avoid
  non-standard packages; *everything* (tables, captions, Methods, refs) in a **single .tex file**;
  graphics supplied separately, not called at compile time; numerical citation style only, in text
  order; BibTeX users must paste the `.bbl` content into the `.tex` and delete `\bibliography`;
  must compile clean with no errors/warnings; **upload a PDF alongside**.
- **Order of elements (exact):** title → authors → affiliations (+present addresses) → bold first
  paragraph → main text → references → tables → figure legends → Methods (+Methods refs, with Data
  and Code Availability at the end) → acknowledgements → funding → author contributions →
  competing interests → additional information (SI line + corresponding author line) → Extended Data
  figure legends and table titles/footnotes.
- **Print figures:** vector with editable layers preferred. Accepted: `.ai`, `.eps`, `.pdf`, `.ps`,
  `.svg` (vector); layered `.psd`/`.tif` (layered art); `.psd`, `.tif`, `.png`, `.jpg` (bitmap);
  `.ppt` if fully editable; ChemDraw `.cdx` for structures. Also acceptable: JPEG 300–600 dpi,
  CorelDraw ≤v8, Word, Excel, PowerPoint. **Never**: Canvas, DeltaGraph, TeX, ChemDraw, SigmaPlot
  (convert to PS/EPS/PDF).
- **Figure specs:** RGB (recommended; CMYK accepted — RGB is auto-converted to CMYK for print) ·
  300–600 dpi for photographic · standard widths **89 mm** (single col) / **183 mm** (double col) /
  120–136 mm (1.5 col), page depth **247 mm** · sans-serif throughout (Helvetica/Arial) · panel
  labels 8pt bold lowercase `a, b, c` · other text max 7pt, min 5pt · line weights 0.25–1 pt at
  final size · don't rasterize or outline text/lines · avoid red/green pairing, prefer green/magenta,
  **no rainbow colour scale** · figure legends <250 words each, starting with a title sentence,
  understandable in isolation.
  *(Note: the initial-submission guide says "90 mm / 180 mm, max height 170 mm, font 5–7pt" — the
  final-submission guide's 89/183/247 is the production truth. Both are quoted here because our
  validator will need to pick one; use the production numbers.)*
- **Extended Data files:** RGB, max 300 ppi, **≤10 MB**, JPEG (preferred) / TIFF / EPS **only**,
  must fit one page (max 18 × 24 cm; tables 8.9 cm or 18 cm wide), filenames
  `CorrespondingAuthorSurname_EDfig1.jpg` / `_EDtable1.jpg`.
- **Forms & declarations page:** author checklist · competing interests declaration · **Licence to
  Publish** (Nature does *not* take copyright for original research — authors grant Springer Nature
  an **exclusive licence to publish**; Reviews/commissioned work get a copyright assignment instead;
  US-government/Crown employees get a non-exclusive licence) · standard structure tables for
  cryo-EM / X-ray / NMR (used as Extended Data tables) · reprints ordering.
- **Preprint policy:** original submitted version may be posted any time; **accepted version after
  a 6-month embargo**; the published copyedited version may never be posted.
- **Then:** subediting (copyedit for a general readership + house style) → edited text sent to
  author for approval → e-proof (figures + Extended Data) → author collates one set of corrections
  from all co-authors → sign-off → schedule. Nature reserves final say on style and figure size.
- **Open access:** hybrid/transformative — gold OA or subscription, author's choice.
  **APCs €2,490 (Scientific Reports) → €10,850 (Nature).** Waiver fund for corresponding authors in
  World Bank LIC/LMIC countries (auto-applied at accept-in-principle, opt-out available).
- **Matters Arising:** comments on Nature papers published within the past **18 months**; online only.

---

## 4. bioRxiv / medRxiv — the pragmatic preprint model

Operated by openRxiv (non-profit). Free. **Not peer reviewed, not typeset.**

### 4.1 Submission

- Register first. Then either **one PDF** (text + figures + tables) or **a Word file + separate
  figure files** which their engine converts to a single PDF.
- Figure formats the converter handles: JPEG, TIFF, EPS, PowerPoint. **Not compatible: PICT,
  Bitmap, Excel, PSD** — convert first.
- **LaTeX must be converted to PDF before submission** (may additionally be uploaded as supplemental
  material on bioRxiv).
- Title page must carry title, author names + affiliations, abstract (medRxiv states this explicitly).
- Supplementary material as separate files; large datasets go to community databases (PDB, GenBank,
  BioLINCC…), not the server.
- **One subject category only** on bioRxiv (~25 categories) — deliberately, to keep feeds
  non-redundant.
- bioRxiv result type: **New Results / Confirmatory Results / Contradictory Results**.

### 4.2 Screening (the vetting model)

Screened for: plagiarism · non-scientific content · inappropriate article types · **material that
could endanger individual or public health** (explicitly includes dual-use research of concern and
work that could compromise accepted public-health measures on transmission/immunization/therapy).
Manuscripts with prior concerns get extra scrutiny. **Typically live within 72 hours**, longer over
weekends/holidays or if flagged. Server reserves the right to decline anything for any reason.

**Out of scope on bioRxiv** (auto-decline list, worth copying as a rules engine): case reports/series ·
work not premised on modern biology/physiology · clinical veterinary studies · narrative reviews ·
editorials/letters/opinion · tool announcements without data · advertisements · addenda · hypotheses
without new data · lab protocols alone · meeting reports · white papers/policy statements ·
guidelines · theses/dissertations/term papers/textbook excerpts · teaching materials · individual
figures/tables/datasets. Clinical work → medRxiv. Non-English → declined (they can only screen
English). **Images that could identify an individual → declined; blurring/pixelation is not enough.**

medRxiv scope is inverted for the health sciences: original research, systematic reviews and
meta-analyses, data articles, methodological research, clinical study protocols — **yes**; narrative
reviews, hypotheses, case reports, editorials/commentary/opinion, correspondence, standalone lab
protocols — **no**.

### 4.3 medRxiv's extra declarations (the health-sciences compliance block)

At submission the author must declare: all relevant ethical guidelines followed · IRB/ethics
committee approvals obtained · patient/participant consent obtained and forms archived · clinical
trial ID from an internationally recognized registry · **competing interests including explicit
acknowledgement of any payments or services from a third party in the past 36 months** · funding
statement including third-party payment for any aspect of the work · a data availability statement
with links to externally hosted data.

### 4.4 Lifecycle

- **DOI at posting** (`10.1101/2019.12.11.123456`, date-stamped since 11 Dec 2019). **Revisions keep
  the same DOI**; version-specific URLs (`…v2`) for citing a specific version.
- **Cannot be removed once posted** — it is citable. medRxiv supports a **"Withdrawn"** marking with
  a posted reason; the original stays reachable via Info/History. Removal only for rare legal reasons.
- Licenses offered: CC BY · CC BY-NC · CC BY-ND · CC BY-NC-ND · CC0 · **no reuse**. Author keeps
  copyright. Posting = consent to text mining.
- **B2J / M2J** — direct transfer of files + metadata from the preprint server into ~250 partner
  journals and peer-review services, so authors never re-key anything. This is the single most
  valuable integration in the whole space and the one I'd copy first (see §19).
- Journals can post their peer reviews onto the preprint's dashboard (J2B, review-linking).

---

## 5. PLOS ONE — the "sound science" journal (runs on Editorial Manager)

**Publication criteria (7)** — notably *not* including importance or novelty:
1. Original research (not reviews, case reports, hypothesis papers, letters, commentaries, essays,
   opinion, policy papers, practice guidelines, other secondary literature, monographs). Accepted:
   Registered Reports (protocol + outcome), systematic reviews, methods/software/database/tool
   papers, qualitative research, **negative/null results**, Lab and Study Protocols.
2. Not published elsewhere, not under consideration elsewhere. **Preprints are fine** (bioRxiv,
   arXiv, theses, conference presentations).
3. High technical standard, described in enough detail to reproduce.
4. Conclusions supported by the data.
5. Intelligible standard English — **PLOS ONE does not copyedit**, so language is a rejection ground.
6. Meets ethics standards (human/animal/tissue, non-commercial cell lines, field sampling, biosafety
   — with approval body + permit numbers, or an explanation of why approval wasn't needed). COPE member.
7. Adheres to reporting guidelines and data-availability community standards.

**Manuscript structure:** Title page → Abstract (**≤300 words, no citations**) → Introduction →
Materials and Methods → Results → Discussion → Conclusions (optional) → Acknowledgments → References
→ Supporting information captions. Figure captions after first mention. Double-spaced, standard
fonts (not Symbol), **continuous page and line numbers**, ≤3 heading levels, no footnotes, no
multi-column layout. Files: DOC/DOCX/RTF (unlocked/unprotected); LaTeX → submit as PDF. Figures as
**separate files**, cited in ascending order, filename matching the citation (`Fig1.tif`).
Supporting info prefixed `S` (`S1 Table`). References: **Vancouver/ICMJE**, numbered in square
brackets, first six authors then et al. Cover letter ≤1 page: contribution, relation to prior work,
article type, suggested editors.

**Entered in the system, not the file:** author names + affiliations, corresponding author,
**CRediT contributions per author**, financial disclosure (funders, grant numbers, funder
role), competing interests, ethics statement, data availability statement, originality/ethics
confirmations.

---

## 6. eLife — publish, review, assess (no accept/reject)

1. Submit directly, or transfer from bioRxiv/medRxiv.
2. Editors (working researchers) decide **whether to review**. That is the only gate.
3. Reviewers write **public** reviews.
4. **Reviewed Preprint** published = manuscript + public reviews + **eLife Assessment** + optional
   author response. Citable DOI.
5. Author may post revised versions (assessment updated), and may finally elect a **Version of
   Record** (funder-OA compliant).

**The eLife Assessment vocabulary** — a controlled two-axis rubric, and the single best UI idea in
modern publishing. Copy this pattern.

*Significance of findings:* **Landmark** (profound, widespread influence) · **Fundamental**
(substantially advances major research questions) · **Important** (implications beyond a single
subfield) · **Valuable** (implications for a subfield) · **Useful** (focused importance and scope).

*Strength of support:* **Exceptional** (sets new standards) · **Compelling** (more rigorous than
state-of-the-art) · **Convincing** (validated, in line with state-of-the-art) · **Solid** (broadly
supports the claims, minor weaknesses) · **Incomplete** (claims only partially supported) ·
**Inadequate** (methods/data do not support the primary claims).

Fee: **$3,750**, waived for anyone who cannot pay.

---

## 7. F1000Research — publish-then-review

Published **as soon as it passes prepublication checks** (originality, readability, author
eligibility, policy/ethics compliance). Peer review by invited experts happens **openly, after
publication**, and the article stays published regardless of the reports. Articles that pass review
get indexed in Scopus/PubMed; everything is in Google Scholar regardless. Authors respond openly to
reports; revised versions are free.

Pre-submission requirements: work is original (preprints fine) · **at least one author is a
qualified researcher** who made a key contribution · meets research/publication standards · all
methodological details and data available for replication · full author + affiliation info + a
conflict-of-interest statement · agreement to
pay the APC (payer name, email, institution, country; VAT number if EU).

---

## 8. Registered Reports (the format that changes the state machine)

Two-stage, review-before-results:

- **Stage 1 manuscript** = Introduction + Methods + pilot results. Reviewed on the question and the
  design.
- → **In-Principle Acceptance (IPA)**: publication is virtually guaranteed and **cannot be revoked
  based on the outcomes** — only for failing quality assurance, deviating from the registered
  protocol, or unresolvable reporting problems.
- Authors run the study, then submit **Stage 2** = the same Intro + Methods + Results + Discussion,
  with pre-registered analyses separated from a clearly labelled "Exploratory Analyses" section;
  data/scripts deposited (OSF, Figshare).
- 300+ journals participate. The Stage 1 protocol can be registered on the OSF Registry.

Implementation consequence: your Submission entity needs a `stage` field and a decision type
`in_principle_acceptance` that locks the methods section.

---

## 9. The three manuscript systems (what authors actually click)

### 9.1 ScholarOne Manuscripts — "up to 7 steps"

Author Centre → **Begin Submission**. Red asterisk = required; `?` tooltips everywhere.
Typical step order: **Type/Title/Abstract → Attributes (keywords) → Authors & Institutions →
Reviewers & Editors (preferred/non-preferred) → Details & Comments (funding, confirmations, cover
letter) → File Upload → Review & Submit.**

- **Keywords**: journals with a controlled list get type-ahead + "add".
- **Funding**: "Add Funder", picked from the **FundRef/Crossref Funder Registry**.
- **Co-authors**: searched by email; if not found, you create the account for them; drag to reorder;
  tick to confirm the list matches the title page.
- **Files**: each gets a **file designation** from a per-journal dropdown; **everything designated
  "for review" is merged into one PDF proof** whose page order you control; anything marked
  **"file not for review"** is excluded from that proof (this is how the de-anonymized full version
  is carried alongside a double-anonymized submission).
- **Review & submit**: red crosses mark missing info; you must open the PDF proof; on double-
  anonymized journals, an un-anonymized proof gets the paper returned.
- **Revisions**: from the Author Centre, "Create a revision" (never a new submission), with a
  visible **days-left-to-revise** counter; you must "view and respond to the decision letter" and
  paste a point-by-point response; ID gets `.R1` appended (`TANDF-2010-0014.R1`); drafts sit in a
  "Revised Manuscripts in Draft" queue.
- Unsubmitted drafts live in an "Unsubmitted Manuscripts" queue, highlighted in blue.

### 9.2 Editorial Manager — the status vocabulary authors obsess over

`Submitted to Journal` → `With Editor` (an editor is assigned/doing checks; recurs at every editor
hand-off) → `Reviewers Invited` → `Under Review` → `Required Reviews Complete` → `Decision in
Process` (typically 1–3 weeks) → `Decision Sent` / `Revise` → `Accept` → `Transfer to Production`.

Design lesson: **authors read the status string as a promise.** Whatever vocabulary we choose, it
must be documented on a public page, and every transition must write an AuditEvent with a timestamp
so "how long has it been in X" is answerable.

### 9.3 Open Journal Systems (OJS) — the open-source reference architecture

Four stages: **Submission → Review → Copyediting → Production.**

- **Dashboard queues:** My Queue (My Assigned) · Unassigned · All Active · Archives (declined +
  published; deleting a declined submission purges its files). Filters: status
  (incomplete/overdue/inactive 30 days), stage, section, assigned editor(s), assigned issue.
  Multiple values inside one filter = OR; across filters = AND.
- **Auto-assignment on submission** by section or category; notification copies can go to the
  journal's primary contact.
- **Submission record panels:** Submission Files · **Pre-Review Discussions** (editor↔author,
  editor↔editor) · Action buttons · Participants · **Submission Library** (COI forms, galley
  approval forms — visible to all participants *except reviewers*) · Document Library
  (journal-wide templates) · Preview · Metadata.
- **Anonymity is enforced at download time**: reviewer downloads are renamed
  `ojs-review-assignment-1-article-text-8.docx`.
- **Pre-review actions:** *Send to Review* · *Accept and Skip Review* · *Decline Submission*.
  (Buttons don't appear until an editor is assigned — a good guardrail.)
- Recorded decisions replace the buttons and can be reverted with **Change Decision**.
- **Granular metadata permissions:** global (`Permit submission metadata edit` on the Author role)
  or per-submission/per-stage (`Allow this person to edit publication details`), which unlocks
  Title & Abstract, Contributors, and Metadata on the Publication tab.
- Review stage: multiple **rounds**, review forms, reviewer recommendations, then Request Revisions
  (with or without a new round) / Accept / Decline / Send to Copyediting.
- Copyediting: assign copyeditor → author replies → final copyedited file → notify editor →
  Production: **galleys** (PDF/HTML/XML/EPUB) → schedule into an issue → publish.

---

## 10. Master field inventory — the complete submission form

This is the union of every field the above systems collect. Mark each as required / optional /
conditional per venue-config; that's how one form engine serves all three archetypes.

**A. Submission basics**
`article_type` (research / review / methods / registered report S1 / S2 / data note / comment) ·
`section` or `subject_category` (primary + optional cross-lists) · `title` · `running_title` ·
`abstract` (plain or structured) · `keywords` (free or controlled vocabulary) · `language` ·
`word_count` · `figure_count` · `table_count` · `reference_count` · `result_type`
(new/confirmatory/contradictory) · `is_resubmission` + prior ID · `related_manuscripts[]` (title,
journal, status, uploaded copy).

**B. People**
For each contributor: `given_name` · `family_name` · `suffix` · `email` · `affiliations[]`
(institution, **ROR ID**, dept, city, country) · `is_corresponding` ·
`is_co_first` / equal-contribution flag · `present_address` · `credit_roles[]` (the 14 CRediT roles,
§13) · `author_order`. Plus: `group_or_collaboration_name` · `non_author_contributors[]` for
acknowledgements (with written permission recorded) · `deceased` flag.

**C. Reviewer preferences**
`suggested_reviewers[]` (name, email, institution, reason) · `opposed_reviewers[]` (max 1–2 groups
at Nature; require a reason) · `preferred_editor` · `review_model_choice` (single/double-anonymized,
opt-in transparent review).

**D. Declarations** *(each needs a stored answer, a free-text field, and a timestamp)*
competing interests (incl. 36-month third-party payments) · funding (funder from the Crossref Funder
Registry + grant numbers + funder role in the study) · ethics approval (body, protocol number, or
justification for exemption) · informed consent (obtained, forms archived) · consent for
identifiable images · animal welfare / ARRIVE · clinical trial registration (registry, ID,
registered before enrollment) · biosafety / dual-use research of concern · data availability
(statement type + repository + accession/DOI + access restrictions) · code availability ·
materials availability · **AI/LLM use disclosure** (which tool, what for, where documented) ·
preprint status + preprint DOI · prior presentation/thesis · plagiarism/originality attestation ·
"all authors have seen and approved" attestation · authorship-change requests (require signed
agreement from **all** listed authors plus the person added/removed).

**E. Files** — see §12.

**F. Rights**
`license` (from the venue's allowed set) · `copyright_holder` · signed **licence-to-publish** or
copyright-transfer or government/Crown variant · `oa_route` (gold / green / subscription) ·
`apc_payer` (name, email, institution, country, VAT) · waiver request + eligibility basis ·
`embargo` for green OA.

**G. Compliance checklists** — reporting summary, CONSORT/PRISMA/ARRIVE/STROBE/REMARK/BRISQ as
applicable, plus a per-venue author checklist.

---

## 11. Peer review models — pick one per venue, per section

| Model | Author sees reviewer? | Reviewer sees author? | Reports public? | Notes |
|---|---|---|---|---|
| Single-anonymized | ✗ | ✔ | ✗ | Industry default. |
| Double-anonymized | ✗ | ✗ | ✗ | Requires an anonymized manuscript + a "not for review" full version; **anonymization is the author's job** and the system must show them the exact proof reviewers will see. |
| Triple-anonymized | ✗ | ✗ | ✗ | Editor also blinded. Rare. |
| Transparent | ✗ (unless signed) | ✔ | ✔ **on publication**, opt-in | Nature's model: reports + rebuttals as a supplementary file. |
| Open / public | ✔ | ✔ | ✔ | F1000Research, eLife public reviews. Reviewers named. |
| Post-publication | n/a | ✔ | ✔ | F1000Research; article published first. |
| Collaborative / consultative | varies | ✔ | sometimes | Reviewers see each other's reports and confer (eLife, EMBO). |
| Registered Report | ✗ | ✔ | ✔ (protocol) | Review before results (§8). |

**ReviewReport schema:** `recommendation` (enum) · `comments_to_author` (rich text) ·
`confidential_comments_to_editor` · structured ratings (Nature's six axes in §3.4, or eLife's
two-axis vocabulary) · `competing_interests_declared` · `signed` (bool → reviewer name published) ·
`consent_to_publish_report` · `consent_to_be_named` · `time_spent` · `attached_annotated_file` ·
`ai_use_declaration` (COPE now explicitly treats undisclosed AI-drafted reviews as an integrity
issue) · `due_date`, `reminded_at[]`, `submitted_at`.

**Recommendation enum** (superset): `accept` · `minor_revision` · `major_revision` ·
`reject_and_resubmit` · `reject` · `transfer_to_sister_journal` · `unable_to_review` ·
`decline_conflict`.

---

## 12. File formats — the complete matrix

### 12.1 Manuscript body

| Format | arXiv | Nature initial | Nature final | bioRxiv | PLOS ONE | Our engine |
|---|---|---|---|---|---|---|
| LaTeX source | **preferred** | ✗ | ✔ (single .tex, converted to Word) | ✗ (as supplement only) | ✗ (PDF instead) | **native** |
| PDF | ✔ (if not TeX-derived) | ✔ (≤30 MB, line-numbered) | alongside .tex | ✔ | ✔ (for LaTeX) | output |
| DOCX | ✗ | ✔ (auto line-numbered) | **preferred** | ✔ (converted) | ✔ | import target |
| RTF | ✗ | ✗ | ✗ | ✗ | ✔ | – |
| HTML | ✔ (+JPEG/PNG/GIF) | ✗ | ✗ | ✗ | ✗ | – |
| DVI / PS | ✗ | ✗ | ✗ | ✗ | ✗ | – |
| Scanned | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ |

### 12.2 Figures

- **Vector (preferred for line art):** AI, EPS, PDF, PS, SVG.
- **Raster:** TIFF/PSD (layered, editable) at **300–600 dpi**; PNG; JPEG (highest quality setting).
- **Specialist:** ChemDraw `.cdx` (structures), `.cif`/`.hkl`/`.fcf` (crystallography), PDB/EMDB
  accessions (structures), `.ppt` only if fully editable and unstyled.
- **Never accepted anywhere:** Canvas, DeltaGraph, SigmaPlot, PICT, Bitmap, Excel-as-figure, PSD to
  bioRxiv's converter.
- Colour: RGB recommended (wider gamut, faithful fluorescence online); CMYK for print fidelity;
  RGB→CMYK conversion mutes bright colours and reduces contrast.
- Accessibility: avoid red/green contrast pairs; prefer green/magenta or another colour-safe palette;
  **no rainbow scales**.

### 12.3 Supplementary & data

Supplementary files (video/audio/large tables/code) as separate uploads with size caps
(Nature Extended Data **≤10 MB per file**). **Large datasets belong in domain repositories**, not on
the server: GenBank/ENA/DDBJ (sequence), PDB/EMDB (structures), SRA, GEO/ArrayExpress, ClinicalTrials.gov,
Dryad, Figshare, Zenodo, OSF, protocols.io (methods). Accession numbers go in the manuscript and must
be **publicly released on the publication date, not later**.

### 12.4 File designations (the dropdown that runs everything)

`manuscript_main` · `manuscript_anonymized` · `title_page` · `cover_letter` · `figure` ·
`figure_source` · `table` · `supplementary` · `data_file` · `code` · `reporting_summary` ·
`checklist` · `ethics_document` · `permission_letter` · `response_to_reviewers` ·
`related_manuscript` · `not_for_review` · `production_source`.

Each designation carries booleans: `include_in_review_pdf`, `visible_to_reviewers`,
`visible_to_authors`, `goes_to_production`, `must_be_anonymized`.

---

## 13. Authorship, contributorship, ethics

**ICMJE's 4 authorship criteria** — all four must be met, by everyone listed:
1. Substantial contributions to conception/design **or** acquisition/analysis/interpretation of data;
   **AND**
2. Drafting or critically revising for important intellectual content; **AND**
3. Final approval of the version to be published; **AND**
4. Agreement to be accountable for all aspects of the work.

Anyone meeting fewer than all four → acknowledgements, not authorship (funding acquisition alone,
general supervision alone, writing/technical/language editing alone do not qualify). **Editors do not
arbitrate authorship disputes** — the institution does. Author add/remove after submission requires a
signed statement of agreement from *all* listed authors and the person being added or removed. ICMJE
also flags exclusion of local LMIC researchers from authorship as grounds for questioning and possibly
rejection.

**AI:** disclose at submission whether AI-assisted technologies were used, describe the use in the
cover letter and in the manuscript (acknowledgements for writing assistance; Methods for data
collection/analysis/figure generation). **Chatbots cannot be authors or be cited as authors.** Humans
remain responsible for everything, including plagiarism in AI-generated text and images. arXiv
requires disclosure of significant generative-AI use as part of methodology standards.

**CRediT — the 14 roles** (ANSI/NISO standard, CC BY 4.0, so we can embed it verbatim):
Conceptualization · Methodology · Software · Validation · Formal analysis · Investigation ·
Resources · Data curation · Writing – original draft · Writing – review & editing · Visualization ·
Supervision · Project administration · Funding acquisition.

**COPE** (Committee on Publication Ethics) provides the flowcharts we should implement as editor
runbooks: ghost/guest/gift authorship · undisclosed COI in a submitted manuscript · peer-review
manipulation suspected during review · concerns raised directly or via social media · removing an
author after publication · risk in published data.

**Corrections taxonomy** (build these as first-class publication events, each with its own DOI and a
bidirectional link to the original):
- **Correction / Erratum** (publisher error) and **Corrigendum** (author error) — minor.
- **Expression of Concern** — when the editor is uncertain and cannot yet resolve it (insufficient
  info, delays, uncooperative parties). *Publish this rather than delay.*
- **Retraction** — when the editor no longer has confidence in the results/conclusions: major errors,
  data/image irregularities, misrepresentation (fraud, identity theft, fictitious authorship),
  redundant publication without attribution, unauthorized use of material/data, unethical research,
  compromised peer review, or undisclosed COI that biases the work. Purpose is **to correct the
  literature, not to punish**. Notices must link to the retracted article, name title + authors, be
  published promptly, and be **free to read**. The **editor** makes the final call (editorial
  independence), the publisher supports. **Batch retractions** are a defined mechanism for paper-mill
  contamination.
- **Removal** — legal only, extremely rare.

---

## 14. Post-acceptance: production and the persistent record

1. **Copyediting** — language, house style, clarity for non-specialists. (PLOS ONE skips this and
   rejects on language instead — a legitimate cost-saving choice for us.)
2. **Author approval of the edited text**, before typesetting.
3. **Typesetting → galleys/proofs**: PDF + HTML + **JATS XML**. Nature cycles e-proofs between
   production, typesetter and author; **one collated correction set from the corresponding author**.
4. **JATS** (ANSI/NISO Z39.96, *Journal Article Tag Suite*) is the archival XML format —
   the Journal Publishing Tag Set is the one to target. This is what PubMed Central, Portico,
   CLOCKSS and most indexers ingest. If we ever want to be indexed, we produce JATS.
5. **DOI registration with Crossref** (deposit XML, UTF-8, numeric character entities only — named
   HTML entities like `&eacute;` are invalid; upload via the admin tool or HTTPS POST). Metadata
   segments Crossref documents and expects: **titles, contributors, affiliations (with
   ROR), abstracts, full-text URLs, license information, funding information, references, ISSN/ISBN,
   article IDs, archive locations, relationships (preprint↔VOR, correction↔original), MathML,
   multi-language content.**
6. **Deposit references** (enables cited-by counts) and **register the preprint→journal-article
   relationship** so the preprint DOI points at the published version.
7. **Preservation:** CLOCKSS / Portico / LOCKSS — required by DOAJ-grade journals.
8. **Indexing applications:** DOAJ → Scopus → PubMed/PMC → Web of Science.

---

## 15. What it takes to be a *real* journal (DOAJ's checklist, condensed)

DOAJ inclusion is the de-facto legitimacy bar for a new OA journal. Requirements:

- **Publishing volume**: ≥5 research articles/year; a new journal needs >1 year of history **or** ≥10
  published OA research articles before applying.
- **ISSN** registered and confirmed at issn.org, displayed on the site; the title must match exactly.
- **Own dedicated URL**; **one unique URL per article**; full text in HTML or PDF minimum.
- **Immediate OA**: no embargo, no registration wall, open license (CC or equivalent).
- **Publicly posted, from the homepage:** open access policy · aims and scope · editorial board with
  **institutional affiliations of every member** · instructions for authors · editorial process
  (peer review) · licensing terms · copyright terms · **all author charges** (submission fees,
  editorial processing charges, APCs, page charges, colour charges, waiver conditions; withdrawal
  charges may not exceed author charges; "no charges" must be stated explicitly) · publisher name,
  address, contact person, dedicated journal email.
- **Quality control:** an editor **and** an editorial board (recommended ≥5 editors, not all from one
  institution, listed with names + affiliations); **≥2 independent reviewers per article** (arts &
  humanities may substitute editorial review by ≥2 editors); peer-review type documented, including
  for special issues.
- **Endogeny ≤25%** — papers where an author is an editor/board member/reviewer, measured over the
  last two issues (or the last calendar year for continuous publishing).
- **Special issues:** EiC responsible for all content; guest-edited papers get the same external
  review and are clearly labelled; guest editors' credentials checked; guest editors' own papers
  independently handled and ≤25% of the issue.
- Recommended: plagiarism checking (Crossref Similarity Check / iThenticate); display of
  submitted / accepted / published dates on each article.
- **No Impact Factor imagery from unofficial services**; DOAJ is a DORA signatory.
- No intrusive advertising.

---

## 16. Screening & integrity automation (what to build into the pipeline)

Ordered by cost/benefit for us:

1. **Similarity check** — Crossref Similarity Check (iThenticate) or an open alternative. Threshold +
   human review of the report, never an automatic reject.
2. **Scope/type classifier** — bioRxiv's out-of-scope list (§4.2) as rules + an LLM classifier that
   proposes a category and flags out-of-scope types. We already have paper-classification
   infrastructure in this product.
3. **Metadata completeness validator** — required fields, ROR resolution, funder
   registry match, DOI syntax, ISSN, trial ID resolves in a registry.
4. **Compile/render gate** — we already have this: Tectonic in E2B. A submission that doesn't build
   isn't a submission (arXiv's model exactly).
5. **Figure integrity** — duplicate/spliced-band detection, EXIF/manipulation checks, minimum dpi,
   **identifiable-person detection** (bioRxiv declines these outright; blurring is not sufficient).
6. **Reference integrity** — do the cited DOIs resolve? Are any of them retracted? (Retraction Watch
   data is available via Crossref.) Are the references real? (Fabricated-citation detection is now a
   first-order need in an AI-drafted world.)
7. **Author/affiliation sanity** — disposable email domains, paper-mill patterns, sudden
   authorship-order churn, reviewer-suggestion emails on free domains (the classic peer-review
   manipulation signature COPE has a flowchart for).
8. **AI-text signals** — flag for human review only. Never auto-reject.
9. **Dual-use / biosecurity screen** — required if we ever touch health sciences.

---

## 17. Decision & status vocabulary (proposed)

```
DRAFT → SUBMITTED → SCREENING ──┬─→ RETURNED_TO_AUTHOR (fixable: format, missing declaration)
                                ├─→ DECLINED_SCOPE / DECLINED_ETHICS / DECLINED_INTEGRITY
                                └─→ SCREENED_OK
SCREENED_OK → (archetype A) POSTED
            → (archetype C) PUBLISHED_AWAITING_REVIEW
            → (archetype B) WITH_EDITOR → EDITOR_DESK_REJECT
                                        → REVIEWERS_INVITED → UNDER_REVIEW
                                          → REVIEWS_COMPLETE → DECISION_PENDING → decision
decision ∈ { ACCEPT, ACCEPT_WITH_MINOR_REVISIONS, MAJOR_REVISION,
             REJECT_RESUBMISSION_INVITED, REJECT, TRANSFER_OFFERED,
             IN_PRINCIPLE_ACCEPTANCE (registered reports) }
ACCEPT → IN_PRODUCTION → PROOFING → PUBLISHED
PUBLISHED → { CORRECTED, EXPRESSION_OF_CONCERN, RETRACTED, WITHDRAWN }
Any state → APPEAL_OPEN → APPEAL_UPHELD | APPEAL_DENIED (final)
```

Every transition writes `{actor, from, to, at, reason, visibility}`. Author-visible status strings
must be documented publicly — see §9.2.

---

## 18. Timelines observed in the wild (for SLA design)

| Step | arXiv | bioRxiv/medRxiv | Nature | F1000Research |
|---|---|---|---|---|
| Screening/moderation | same-day (14:00 ET cut-off → 20:00 ET live) | usually **≤72 h** | – | days |
| Desk decision | – | – | **~1 week** if not going to review | – |
| Referee turnaround | – | – | **~2 weeks** agreed | post-publication, open-ended |
| Decision after reports | – | – | "routinely very rapid" | – |
| Rounds to acceptance | – | – | **1–3, sometimes more** | unlimited revisions |
| "Decision in Process" (EM) | – | – | 1–3 weeks industry-wide | – |

---

## 19. Recommendation for ScholarXIV

### 19.1 What we already have that matters

`src/lib/write/` is most of the submission client: flat-path project files, `rootFile`, text/binary
split (Mongo + S3), Tectonic-in-E2B compile with parsed `Problem[]` diagnostics, a PDF + page-one
thumbnail, arXiv import (`arxiv.ts`, `tar.ts`), and templates. **A submission is a frozen snapshot of
a `Project` plus a metadata record.** That is a small step from where the code is.

### 19.2 Phased plan

**Phase 1 — "ScholarXIV Preprints" (archetype A).** Submit from `/write` in one click: freeze the
project (immutable file snapshot + compiled PDF), collect the metadata form, pick a license, agree to
terms, land in a moderation queue. Moderator console with accept / reclassify / return-for-fixes /
decline + reason. Public landing page per preprint, versioned (v1, v2…), DOI at posting. Everything
here is directly modelled on arXiv + bioRxiv, both of which are documented above field-for-field.

**Phase 2 — Review layer (archetype C).** Invite reviewers, public reports, and an **assessment
rubric copying eLife's two-axis controlled vocabulary** (significance × strength of evidence). This
gives us a differentiated product without ever having to say "reject" — and without needing an
editorial board on day one.

**Phase 3 — Full journal (archetype B).** Editorial roles, review rounds, decision letters,
point-by-point response UI, copyediting/production, issues, JATS output, DOAJ application.

**Phase 4 — Interop.** The highest-leverage feature in this entire dossier is bioRxiv's **B2J**:
push files + metadata from us into an external journal (and pull from arXiv/bioRxiv into us) so
authors never re-key. We already have arXiv import; the reverse direction is the moat.

### 19.3 Decisions — SETTLED (2026-07-26)

| # | Decision | Answer |
|---|---|---|
| 1 | **Archetype** | **A — preprint server, arXiv-style.** No review process. Every submission is listed automatically; no human moderation gate for now. |
| 2 | **DOIs** | Open. See [scholar-profiles-and-indexing-research.md](scholar-profiles-and-indexing-research.md) §5 — DataCite (what arXiv uses, prefix `10.48550`) is the cheap path for preprints. |
| 3 | **Permanence** | Open. Follows from (2). |
| 4 | **Scope** | **All disciplines.** |
| 5 | **Money** | **Free.** |
| 6 | **Input format** | **LaTeX source required.** PDF/DOCX policy analysed in [scholar-profiles-and-indexing-research.md](scholar-profiles-and-indexing-research.md) §1. |

Consequence of (1): with no human moderation, the automated screening layer in §16 stops being a
nice-to-have. See the de-indexing risk in the companion doc, §2.6.

---

## Sources

arXiv: [submission process](https://info.arxiv.org/help/submit/index.html) ·
[metadata prep](https://info.arxiv.org/help/prep.html) ·
[licenses](https://info.arxiv.org/help/license/index.html) ·
[sizes](https://info.arxiv.org/help/sizes.html) ·
[endorsement](https://info.arxiv.org/help/endorsement.html) ·
[moderation](https://info.arxiv.org/help/moderation/index.html) ·
[withdrawals](https://info.arxiv.org/help/withdraw.html) —
Nature: [initial submission](https://www.nature.com/nature/for-authors/initial-submission) ·
[formatting guide](https://www.nature.com/nature/for-authors/formatting-guide) ·
[editorial criteria and processes](https://www.nature.com/nature/for-authors/editorial-criteria-and-processes) ·
[final submission](https://www.nature.com/nature/for-authors/final-submission) ·
[forms and declarations](https://www.nature.com/nature/for-authors/forms-and-declarations) ·
[peer review policy](https://www.nature.com/nature-portfolio/editorial-policies/peer-review) ·
[open access](https://www.nature.com/nature-portfolio/open-access) —
[bioRxiv submission guide](https://www.biorxiv.org/submit-a-manuscript) ·
[bioRxiv FAQ](https://www.biorxiv.org/about/FAQ) ·
[medRxiv submission guide](https://www.medrxiv.org/submit-a-manuscript) —
[PLOS ONE submission guidelines](https://journals.plos.org/plosone/s/submission-guidelines) ·
[PLOS ONE criteria for publication](https://journals.plos.org/plosone/s/criteria-for-publication) —
[eLife peer review model](https://elifesciences.org/about/peer-review) ·
[eLife assessments](https://elifesciences.org/inside-elife/db24dd46/the-elife-model-what-is-an-elife-assessment) —
[F1000Research how to publish](https://f1000research.com/for-authors/publish-your-research) —
[ScholarOne author guide (T&F)](https://authorservices.taylorandfrancis.com/publishing-your-research/making-your-submission/using-scholarone-manuscripts/) ·
[Editorial Manager statuses (Wolters Kluwer)](https://www.wolterskluwer.com/en/expert-insights/authors-decoding-your-manuscript-status) ·
[OJS editorial workflow](https://docs.pkp.sfu.ca/learning-ojs/3.3/en/editorial-workflow) —
[ICMJE authorship](https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html) ·
[CRediT](https://credit.niso.org/) ·
[COPE flowcharts](https://publicationethics.org/guidance/Flowcharts) ·
[COPE retraction guidelines](https://publicationethics.org/guidance/guideline/retraction-guidelines) —
[Registered Reports (COS)](https://www.cos.io/initiatives/registered-reports) —
[Crossref XML deposit](https://www.crossref.org/documentation/register-maintain-records/direct-deposit-xml/) ·
[Crossref metadata segments](https://www.crossref.org/documentation/schema-library/markup-guide-metadata-segments/) ·
[JATS tag library](https://jats.nlm.nih.gov/publishing/tag-library/1.3/chapter/getting-started.html) ·
[DOAJ guide to applying](https://doaj.org/apply/guide/)

## Sitemap

See the full [sitemap](/sitemap.md) for all pages.
Docs-scoped sitemap: [/docs/sitemap.md](/docs/sitemap.md).
Well-known sitemap: [/.well-known/sitemap.md](/.well-known/sitemap.md).