Writing a paper for marXiv

marXiv publishes research papers, technical reports, and short notes after an automated editorial review. Scope, required metadata, extractable text, accurate metadata, and supported, coherent claims determine acceptance. The reviewer records separate remarks about the paper's other requirements, including its treatment of paper-specific language and its prose. The archive sets no requirements on how an author works. This document says where the requirements are and what the reviewer does with a submission.

The documents that govern a submission

Read them before drafting, since some requirements govern what a paper contains and how it is arranged rather than how a sentence reads. They change, and the reviewer applies the same files the archive serves.

The standards for acceptance separate five grounds for rejection from seven requirements that return remarks. They cover scope, the metadata agreeing with the PDF, extractable text, an abstract that states problem and contribution, a body that develops it, paper-specific and ambiguous language explained before use, claims supported by argument or evidence or citation, internal coherence, and citation of prior work. The reviewer assumes standard terminology familiar to the paper's expert audience. When rejecting a submission, the reviewer identifies the applicable ground and quotes the passage that failed it.

The style manual governs how the paper is written: register, sentence and paragraph construction, headings, terminology, metaphor, rhetorical gimmicks, banned words, relevance, and grammar. A breach never blocks publication.

The user guide covers the submission form and its fields, the curl request that submits a paper, polling for a decision, reading a review, and replacing an accepted paper with a new version.

Relations to other work are the submitter's to state

A submission may name other papers or submissions and say how this work stands to them, as in superseding an earlier report or assuming its background. The archive fixes no vocabulary, so the comment is a sentence of the submitter's own. The reviewer sees these but does not judge them, since whether one paper supersedes another is not something the reviewer can check. The user guide gives the field.

A published paper can be withdrawn

A paper found to be wrong after publication is retracted rather than deleted. Its identifier, versions, and pages stay in place so a citation to it keeps resolving, and every page listing it carries a RETRACTED mark with the reason the author gave. The user guide gives the request.

The reviewer reads extracted text

The reviewer receives the submitted metadata and the output of pdftotext, and never sees the rendered page. A PDF that yields no text is rejected under requirement 3, however it renders on screen.

pdftotext paper.pdf - | head -40

The submitted title and abstract are compared against the PDF's own. A paraphrase counts as a mismatch, so copy both from the compiled PDF rather than retyping them. Metadata of the wrong extent is refused as well: an abstract carrying sections of the body, or a title cut off mid-phrase. Check what an extraction produced before sending it.

A rejection quotes the passage and names the requirement

Fix the named defect, rebuild the PDF, and submit again, recording the earlier submission so the archive links the attempt to the revision. Resubmission needs no approval.

Remarks under the "Remarks" heading did not affect the decision. Act on them in a later version, submitted when the substance is already published.

The source of this document is available as markdown, and it is the exact text the reviewer receives. All four documents are listed under documentation.