Editorial policy

0. Two tiers

Pith Preprint has two tiers with two different doors:

• The preprint record (open). Any completed paid review can publish a preprint record. Admission is an integrity screen, not a quality gate: the full review, scorecard, claim ledger, and verification grade attach to the record as public metadata. A paper whose review says "major revisions needed" publishes carrying exactly that judgment. Judgment is displayed, never a door.
• Journal promotion (selective). Promotion from a preprint record into the journal runs the disclosed rubric, the hard floors, and the public candidate window described below. This tier is deliberately strict; selectivity is its point.

0a. Preprint admission (integrity only)

Two passes run at submission, and only integrity-class failures hold a manuscript:

• Deterministic integrity screen (flags, never scores): duplication against Pith's published records and the arXiv corpus; the reference audit (a bibliography where more than 20% of mechanically-judged references fail resolution or mismatch is held as fabricated); format sanity (the upload parses as a paper); abuse.
• Desk screen (one cheap model pass): three 0–10 axes with verbatim quoted evidence: internal coherence, rigor floor (precision, not correctness), professionalism. It can hold only for not-a-paper, abuse, or an axis at noise grade (2 or below) with quoted evidence. A model failure never holds a paper; plumbing is not evidence.

PASS publishes the record immediately with the review attached; no human step. HOLD routes to a human queue with the evidence attached, and every resolution is logged to the public decision ledger. Disagreement with consensus, unfashionable topics, missing novelty, and weak results are never hold reasons.

1. States (journal promotion tier)

The preprint record has its own axis: preprint published, held (human queue), or withdrawn. The journal ladder below operates independently on the same submission:

1. submitted: author opted into the journal lane on a completed peer-review ticket.
2. eligible: rubric score and hard floors pass; paper can open a public candidate window.
3. candidate: public endorsement and challenge window is open (10 days by default).
4. published: Pith Preprint endorsed the paper. A permanent article page is created.
5. declined: publication was declined; the decision letter is published with the rubric.
6. ineligible: hard floor failed or rubric score below threshold. Authors see exactly which gates failed. The preprint record is unaffected.
7. withdrawn: author withdrew; the rubric and objection record are kept for the audit log.

2. Hard floors (journal promotion only)

Any one of these blocks journal-promotion eligibility, regardless of score. None of them blocks the preprint record:

• reject_recommendation: Final referee recommendation is 'reject'.
• out_of_scope: Final referee recommendation is 'out_of_scope'.
• uncertain_recommendation: Final referee recommendation is 'uncertain'.
• load_bearing_overclaim: At least one load-bearing claim is overclaimed.
• load_bearing_unsupported: At least one load-bearing claim is unsupported.
• unresolved_major_revisions: Major revisions are open. Submit a point-by-point response (and the revised manuscript if it changed) on this ticket; a referee re-assesses it automatically and the gate reopens once the concerns are resolved.
• unresolved_formalization: Formalization required. Provide the missing Lean artifacts and submit a response on this ticket; a referee re-assesses it automatically.
• unresolved_empirical_artifact: Empirical evidence required. Upload the data or code and submit a response on this ticket; a referee re-assesses it automatically.
• minor_revision_pending_check: Minor revisions are open. Submit a response addressing the open items on this ticket; a referee re-assesses it automatically and the journal gate reopens once they are resolved.
• synthesis_missing: Peer review did not produce a synthesis payload.
• ticket_not_done: Peer review ticket is not in a 'done' state.

Hard floors are deterministic. They are computed from the peer-review synthesis, claim ledger, and verification grade. There is no human override.

3. Rubric (journal promotion)

Each axis is scored in [0, 1] and combined by the disclosed weight vector. The composite score is also in [0, 1].

Eligibility threshold: composite score ≥ 0.55. Scores open the journal lane; they never publish a paper by themselves. Publication requires explicit action.

4. Journal fee

There is no fee to publish in Pith Preprint. The only cost in the pipeline is the Pith Review fee, which pays for the machine review that produces the rubric score and reference audit. Once a paper passes review and clears the candidate window, journal publication is free.

5. Endorsements

During the candidate window, signed-in users can add a typed endorsement:

• technical_correctness: the central arguments check out.
• novelty: this is genuinely new work.
• reproducibility: code, data, or formal proof artifacts are real and usable.
• formal_artifact: a Lean module or other machine-checked artifact backs a load-bearing claim.
• field_significance: the result matters for a field the endorser actually works in.
• expository_value: the paper unifies or clarifies a body of prior work.

An author cannot endorse their own submission. Endorsements never override an objection or a hard floor.

6. Objections

Objections during the candidate window must cite at least one claim ID or section reference with at least 20 characters of explanation. Three severities:

• minor: clarification, typo, or small scope concern.
• major: substantive concern about a load-bearing argument.
• blocking: an unresolved objection blocks publication when the window closes.

Authors can respond and mark an objection as resolved. The final decision letter cites every unresolved blocking objection.

7. Models and AI instructions

The rubric is the AI instructions. There is no hidden model prompt. Every score is a deterministic function of:

1. the structured fields in the existing peer-review synthesis JSON;
2. the claim ledger and verification grade in that same JSON;
3. the public endorsement and objection record once the candidate window has run.

The live model roster, read from the deployed configuration (this table cannot drift from what actually runs):

Referee seats are interchangeable as the state of the art moves; the roster is configuration, not doctrine. Uploaded (private) manuscripts are never sent to models whose providers claim training rights on inputs; that exclusion is enforced in the pipeline, not just stated here.

8. Anti-club commitments

• No acceptance based purely on endorsement volume.
• No private editorial veto. The composite rubric, the public endorsement record, and the public objection record together produce the decision.
• Endorsements are weighted by verified identity, never by institutional prestige.
• Authors are anonymous in the candidate window for uploaded manuscripts until they opt into identity disclosure, unless the work is already public on arXiv.
• Every published paper has a public decision ledger reproducible from the rubric version recorded on the page.
• Appeals must target a concrete failed gate or a factual error in the review. Appeals are public.

9. Standing audit

The audit dashboard publishes acceptance rates, score distributions, and objection outcomes. It updates in real time and is the same data feed that drives this rubric.

10. DOI registration

Papers submitted directly through Pith and published in the journal are registered with Crossref as pre-prints (record type posted_content type="preprint") under the DOI prefix 10.67347, held by the Recognition Physics Research Institute.

• When. The DOI is minted at publication, only after the permanent article page is live. Nothing is registered for submissions in candidate, eligible, or any earlier state.
• What it points to. The DOI resolves to the permanent article page at /journal/<slug>, which carries the full text, the review record, the references, and the decision log.
• Pattern. 10.67347/pith.preprint.<year>.<n>, where n is a sequential serial assigned once per submission and never reused.
• Scope. Only directly-submitted pre-prints. Papers reviewed from arXiv keep their arXiv identity; papers that already carry a publisher DOI keep it. Pith never mints a duplicate identifier.
• Status. Crossref registration is asynchronous. Until Crossref confirms, the article page labels the DOI with its registration state; once confirmed, the label is dropped.