Submission guidelines

Form

• A manuscript should look like a paper: title, authors, abstract, body. Upload LaTeX source or a finished PDF so PreXiv can host the paper; an external URL (e.g., arXiv, GitHub, journal site, or your homepage) is a supplemental link, not a replacement for the hosted artifact.
• Use the authors line for humans or organizations that can take responsibility for the work. Disclose AI tools in the AI model field, not as legal authors.
• Categorize honestly. misc is fine if nothing fits.

Honesty

• State what role the AI played and what role you played. The conductor notes field is the right place.
• Treat conductor and auditor labels as provenance disclosures. They do not decide legal authorship, copyright ownership, or whether readers may rely on the work for professional decisions.
• Do not claim an auditor who has not actually read the manuscript and signed off in writing. This is the one thing that will get a submission removed.
• Do not pretend the work is more polished than it is. This is wrong somewhere and I want help finding it is a valid reason to post.

Rights and sensitive information

• Submit only material you have authority to publish and license for PreXiv hosting.
• Do not upload confidential datasets, patient information, private addresses, identity documents, trade secrets, export-controlled material, or other restricted content unless you have a lawful basis and every required consent.
• Use the private conductor/model controls before upload when you want identifying details redacted from the public source and PDF. PreXiv can redact LaTeX/source submissions; direct PDF uploads are rejected when private conductor/model fields would need redaction.
• PreXiv content is not legal, medical, financial, engineering, security, or clinical advice. Readers should treat it as unaudited research discussion unless an independent qualified review says otherwise.

Quality

• Edit before posting. Manuscripts that are obviously a single chat turn pasted in will be downvoted.
• Math should typeset. Use $…$ and $$…$$ in the abstract and comments; for the body, the PDF is your responsibility.
• Reproducibility, even partial, helps. Linking a colab / repo / data is a good signal.

Discussion

• Engage with the manuscript, not the conductor. This is wrong because of X is welcome; amateurs shouldn't post is not.
• If you are an expert and you've read it, leaving a signed opinion in the comments is genuinely useful even if you are not the formal auditor.

Withdrawal

Submitters can withdraw their own manuscripts from the manuscript page. The page is kept as a tombstone with the id, DOI, title, original conductor, and withdrawal reason so citations do not break. Admins can also withdraw or permanently delete spam and abuse.