Docs / Editors, referees and their decisions
Editors, referees and their decisions
Every new problem passes through the same stages. This page says, for each stage, what starts it, who acts, what they can decide, what is signed and recorded, who can see what, and how a decision can be challenged. Editors decide on results (solutions, disproofs, progress) the same way: the stages of a result. Referees check accepted solutions and disproofs (referees). Editors also triage and decide contests of results and of decisions on them (contests); the party a decision went against may appeal some of them once.
Editors and referees have no deadline: an item waits in the queue until someone acts, and its page shows its stage. Editors are notified in the site of results and contests held by the screening or waiting for a decision, of results withdrawn, of appeals they may decide, and of held or reported comments; referees, of accepted solutions and disproofs they may recommend on; no one is told of what they did themselves (who is told what).
PubPhys has one editor for now; what that editor may not decide waits for another.
Who decides
Any account with a linked ORCID iD, invited or (once ORCID registration opens) registered with its iD, submits problems, revisions and topics through the same screening and review described here; there is no extra hold for new accounts (accounts).
- Editors are appointed for one field each, the chief editor for every field; both are appointed by signed, public records (appointments). Only they decide on problems of their field.
- Referees are appointed for one field each, by the chief editor, the same way. They check accepted solutions and disproofs and recommend them or not (below).
- Authority is bound to the role and checked at the moment the decision is signed: the signer must hold, right then, the appointment the kind of decision needs for the item's field (table below). The decision keeps the appointment it was made under, so a later change of appointments does not undo it.
- Nobody decides on their own problem or result: neither by the account that submitted it nor by the ORCID iD that signed it. Nor does anyone decide on a result on a problem where they have a live result of their own (who is excluded from what).
- The chief editor holds an editor's authority in every field and may decide any problem or result at any time, not only when the field's editors are excluded; the same exclusions apply to the chief editor. If the chief editor is excluded too from an item waiting for an editor (anything but a new problem awaiting its first decision), the item is flagged and the chief editor lends an editor of another field (below).
No eligible editor, and lending
A result, contest, appeal, later revision of a problem or topic revision waiting for an editor is flagged no eligible editor when every editor of its field and the chief editor are excluded from it. A new problem awaiting its first decision is never flagged. Flags are checked hourly and whenever an editor is appointed or revoked; a flag is cleared once someone may decide the item, or when it leaves the queue (decided, withdrawn, removed). The flagged items are listed in the chief editor's queue, and the administrator is told by email.
- The chief editor then appoints an editor of another field to this field: an ordinary signed appointment, public on editors and referees, revocable like any other. Its authority is field-wide, as for every editor; revoking it later does not undo the decisions signed under it.
- An appointment made while the field has flagged items is linked to them. Someone excluded from one of them cannot be appointed to the field until it is settled: the appointment could not decide it.
- Appointing is administrative, so the chief editor appoints even when excluded from the item. When the exclusion is one of interest (the chief editor is the author of the result, of a competing live result, of the contest or of the appeal, or of the problem an appeal concerns), the chief editor must tick an acknowledgement when preparing the appointment, or it is refused (also at signing, if an item was flagged meanwhile), and the appointment's page, its receipt and the list of editors say "appointed while excluded from" the item, by its record hash. This marker is kept by PubPhys next to the signed record, not inside it: the appointment record's format has no place for it yet.
- An appointment linked to a flagged item is lent when its holder was then an editor of another field. A lent editor does not count toward the field's 1 editor (opening a field), and the administrator is told when a lent appointment is still active 14 days after the items it was lent for stopped being flagged. The lent editor is notified in the site (who is told what).
Kinds of decision
| Kind of decision | Decisions | Who may sign it |
|---|---|---|
| Acceptance | accept, reject, not accepted at screening | An editor of the item's field, or the chief editor. |
| Recommendation | recommend, do not recommend | A referee of the item's field. Not an editor as such, and never the chief editor, even one who also holds a referee appointment. |
| Triage | substantive, not substantive | An editor of the field of the contested result, or the chief editor, with no stake in the contest (who decides). |
| Outcome | upheld, dismissed | As for triage; never the editor who triaged the same contest. |
| Credit | credit | As for triage, and only together with the uphold it belongs to. |
| Reversal | reversed | As for triage, and only together with the uphold it belongs to; never the author of the reversed decision. The revert of a cosmetic edit is signed on its own, by an editor of the problem's field or the chief editor, not the problem's author nor the edit's (revisions). |
| Editorial | moved to another topic, status note verified | An editor of the problem's field, or the chief editor; a move to another field's topic only the chief editor. Never the problem's author. |
| Appeal | appeal granted, appeal denied | An editor of the item's field, or the chief editor, who decided nothing in the case before (appeals). An appeal left to the chief editor alone goes to an editor of the field who decided nothing in the case when every chief editor is excluded or suspended, or none is appointed; if the chief editor cannot decide because of an interest of their own that is not yet public, the appeal waits until that interest ends. |
| Written by PubPhys | removed, acceptance carried to a new revision, accept of a seed revision | Not signed by a person: PubPhys writes the record on an editor's instruction (removal) or when it publishes a revision (carried acceptance), and when it publishes a seed revision from the approved seed revision batch or the administrator's seed editor (accept); the audit log, which is internal (only the people who operate PubPhys see it), keeps who gave the instruction. An author's revision whose substance changed, and every proposed or edited topic revision, is decided by an editor. |
| Decision | On what | Meaning | Signed by |
|---|---|---|---|
| Accept | A published problem, result or topic's revision no editor has decided. | A problem: a well-posed open problem in physics a specialist can work on. A result: a serious, relevant contribution of the stated kind; not that it is correct. A topic: a coherent area of the field that open problems sit under (proposing a topic); a topic is decided before the problems under it. | An editor of the field or the chief editor; for a new revision PubPhys publishes whose substance changed, PubPhys itself (revisions). |
| Reject | As for accept. | It is not that; it stays public, marked rejected, and a rejected result stops counting. | As for accept. |
| Not accepted at screening | A held problem, result or topic's revision. | It is never published. The editor picks a category: for a problem not physics, not a problem or unintelligible; for a result off topic, not physics or unintelligible; reasons are optional. A topic's revision takes reasons. | As for accept. |
| Recommend | An accepted, intact, counting solution or disproof. | The referee checked it in full and found it correct and complete (referees). | A referee of the field. |
| Do not recommend | As for recommend. | The referee found an error or a gap, and says where. | A referee of the field. |
| Substantive | A published contest not yet triaged. | It states a specific, checkable objection that would change the result's standing or credit, or the contested decision, if right; it is then open. Not a judgment of who is right. | An editor of the field outside the case. |
| Not substantive | As for substantive. | It does not; the contest is closed and changes nothing. | As for substantive. |
| Upheld | An open contest. | The objection is right: per its grounds the result is refuted or incomplete, earlier work is credited, or the contested decision is reversed (grounds). | An editor of the field outside the case, not the one who triaged it. |
| Dismissed | An open contest. | The objection is not right; nothing changes. | As for upheld. |
| Credit | The contested result. | Names earlier work the result should credit; shown with the result. Signed only with the uphold of a contest on prior art or missing credit. | The editor who signs the uphold. |
| Reversed | A contested acceptance, recommendation or credit, still in force; or a cosmetic edit carried in the last 90 days (a revert). | The decision stops counting; its signer cannot decide the same item again. Signed only with the uphold of the contest against it. A revert puts the revision before the edit back on display. | The editor who signs the uphold; for a revert, an editor of the problem's field. |
| Moved to another topic | The displayed revision of a public problem, or a problem hidden under a rejected or removed topic. | The problem belongs under another accepted topic; its results, comments and followers go with it, and its pending items go to the new field's editors. | An editor of the field; to another field, the chief editor. |
| Status note verified | The displayed revision of a public problem with a status note. | An editor checked the status note against the literature; any later change of the note clears the mark. | An editor of the field or the chief editor. |
| Appeal granted | A pending appeal of a decision still in force. | The appealed decision stops counting (for an uphold, with what came with it) and the item goes back for a fresh decision; a fresh decision of the same kind is final. | An editor with no part in the case. |
| Appeal denied | As for appeal granted. | The decision stands for good. | As for appeal granted. |
| Removed | A problem or result at any stage, or the text of a decision or a withdrawal. | Copyright, personal data, abuse or spam; the content is withheld for good. Also a contest. | PubPhys, on an editor's instruction. |
| Accepted (cosmetic change) | A new revision of an accepted problem that changes the wording only. | The acceptance carries over to the new revision (revisions). | PubPhys, when it publishes the revision. |
A decision on a result targets the result's signed claim record, never the problem; a decision on a problem targets
the problem's signed revision; triage and outcome target the signed contest record; a reversal targets the reversed
decision's record.
The stages of a new problem
| Stage | Starts when | Who acts, and what they can do | Recorded | What they see | Next | How to challenge |
|---|---|---|---|---|---|---|
| Signed | The author signs with ORCID (submitting). | Nobody; the screening starts by itself. | The upload and the signed problem, in the public log. | Everyone: the log entries (that a record exists, not what it says). The author and the field's editors: the content. | Screening, within minutes. | Nothing to challenge yet. |
| Sealed | The author signs the problem sealed: its text stays private until they reveal it. | Only the author, who reveals it (the screening then starts) or abandons it. | The upload and the signed problem, in the public log. | Everyone: the log entries (that a record exists, not what it says). The author alone: the content; no editor sees it. | Screening, once revealed; abandoned, if the author abandons it or deletes the account. | Nothing to challenge yet. |
| Screening | Right after signing. | An automatic check, with the author's consent to AI (what is sent), asks only whether this is an open question in physics, stated intelligibly. It never rejects anything. | Its report. | The author and the field's editors: the problem and the report. | Published if it is confident; otherwise held. Without AI consent, or if AI is off, the problem is held. | Nothing: the screening decides nothing against the author; an editor looks at every held problem. |
| Held | The screening doubted it (for a problem with parents, also whether it relates to them as declared), suspected an attempt to manipulate it, or could not run. | An editor of the field either releases it (published as unreviewed) or signs that it is not accepted at screening, with a category and optional reasons. | Release: in PubPhys's audit log, not signed. Not accepted: a review record signed by the editor. | The author and the field's editors only, including the hold reason and the decision. | Published, unreviewed; or not accepted, and never published. | A "not accepted" decision: appeal by the author within the 30 days, while new problems are accepted; a granted appeal returns it to held. |
| Published, unreviewed | The screening passed it, or an editor released it. | An editor of the field signs accept (a well-posed open problem in physics, stated so that a specialist can work on it, and correctly related to each declared parent) or reject, with reasons. | The decision is a review record signed by the editor. | Everyone: the problem, marked unreviewed. | Accepted or rejected. | Nothing yet; the decision below can be challenged. |
| Accepted | An editor signed accept. | Nobody; the problem is marked accepted. | The signed decision. | Everyone: the problem and the decision, with the editor's ORCID iD and reasons. | Stays accepted. A later revision that changes the wording only carries the acceptance over; one that changes the substance waits for an editor's accept or reject (revisions). | An acceptance is not appealable. |
| Rejected | An editor signed reject. | Nobody. | The signed decision. | Everyone: the problem stays readable, marked rejected, with the editor's ORCID iD and reasons. | Final unless appealed. | Appeal by the author within the 30 days, while new problems are accepted; a granted appeal returns it to published, unreviewed. |
| Removed | An editor removes it for copyright, personal data, abuse or spam, at any stage. | Nobody. | A review record written by PubPhys on the editor's instruction; the editor is in the audit log. | The content and the file are withheld for good; a page of a problem that was public says why; otherwise it is not found. | Final. | Through the administrator only. |
Revisions of an accepted problem
A problem's author revises it, and an editor of the field revises a seed problem, by signing a new revision (revisions). If the precise question is unchanged and only the wording changed (the core fingerprint is the same), PubPhys publishes it with a carry_acceptance record: the new revision is accepted because the old one was, and an editor checks it afterwards and may revert it. If the substance changed, or the edit changes the literature status, the revision waits for an editor's accept or reject. The seed revision batch and the administrator's seed editor publish seed revisions accepted at once: with carry_acceptance for a wording change, otherwise with an accept PubPhys writes. Results keep counting only for the question they answer (what a result freezes).
The queue
An editor's queue lists the held and the undecided published problems and results of the fields they decide (all fields for the chief editor), oldest first, leaving out their own. Revisions have their own sections: held revisions, revisions waiting for a decision (with the diff and the problem's results), and cosmetic edits of the last 90 days to check, which an editor marks checked or reverts. A result on whose problem the editor has a live result of their own is listed, but its page says why they cannot decide it and offers no decision. A referee's queue lists the accepted solutions and disproofs of their fields, oldest first: first those they may referee, then the others with the reason they may not.
How editors and referees sign decisions
- Stage. On an item in the editorial queue the editor chooses a decision and writes the reasons. The browser builds the review record and computes its fingerprint. Staging records nothing and has no effect; a staged decision is kept for a day, and choosing again on the same item replaces it.
- Check. The queue lists the staged decisions with their exact text and fingerprints. The editor ticks the ones to sign (up to 20 at once). At most 20 records of any kind wait to be signed at once for one account; staging more asks to sign or discard some first.
- Sign. One ORCID sign-in, which always asks for the ORCID password, signs all ticked decisions together. It must be finished within 10 minutes; after that, start signing again. Then PubPhys checks every one of them again: the item is still in the stage the decision is for, nobody else decided it meanwhile, the signer still holds the appointment the decision needs for its field and is not excluded. If any check fails, nothing is signed, the page names the decisions that failed, and the editor stages the others again; that signature can never be used later.
- Effect. The decisions take effect at once, go into the public log, are sealed by PubPhys and dated through Bitcoin like every record (how proofs work).
Only one decision of a kind stands on an item: once a problem or a result is accepted or rejected, no second accept or reject can be signed on it. Recommendations are one per referee: a result can have many, but each referee's only once. Editors' and referees' decisions are signed in the same queue and can be signed together.
Referees
An editor's acceptance says that a result is a serious contribution of the stated kind; it does not say that it is correct. Referees check that. Two recommendations from different referees of the field make a problem verified (by a solution) or disproven (by a disproof): how the status follows.
Appointment and opening a field
The chief editor appoints and revokes referees per field by signed records, listed publicly on editors and referees. After the launch, a field takes results only once it has at least 2 editors (the chief editor counts) and 2 referees (the chief editor does not count, even with a referee appointment). An editor counts in one field only, their home field: that of their earliest editor appointment that is not lent (lending). A person may be an editor and a referee of the same field; they then act in one role or the other per result, and the exclusions below decide which.
The stages of refereeing
| Stage | Starts when | Who acts, and what they can do | Recorded | What they see | Next | How to challenge |
|---|---|---|---|---|---|---|
| Awaiting referees | An editor signed accept on a solution or a disproof that is intact (no upheld contest against it) and counts for its problem. | Every referee of the field who is not excluded sees it in the queue, oldest first, reads it in full and stages recommend or do not recommend with reasons. There is no deadline. Progress is not refereed. | Nothing until a referee signs. | Everyone: the result and the editor's acceptance; the problem shows claimed solved or claimed disproven. | Recommendations, one per referee; the result waits for them without a deadline. | The result itself: a contest. |
| Recommend | A referee signs it. | The referee states that they checked the result in full and found it correct and complete for the question as stated. | A review record signed by the referee, keeping the referee appointment and the field it was made under. |
Everyone, with the referee's ORCID iD and reasons. | With a second recommendation from a different referee of the field: verified or disproven, as long as the result counts. | Not appealable. A contest against the recommendation; an upheld one reverses it. |
| Do not recommend | A referee signs it. | The referee states that they found an error or a gap, and says where. | A review record signed by the referee. | Everyone, with the referee's ORCID iD and reasons. | Shown with the result; it blocks nothing by itself and is not counted against the recommendations. | Not appealable. The way to make a result incomplete or refuted is a contest. |
Rules
- Only accepted, intact, counting solutions and disproofs are refereed: not progress, not a result waiting for an editor or rejected, withdrawn, removed, made incomplete or refuted by a contest, nor one that answers an earlier version of the question (it no longer counts). Requiring the result to count, not only to be accepted and intact, is the owner's decision (2026-10-10).
- One recommendation per referee and result, final. A referee cannot change it or sign a second one; only an upheld contest can reverse it, and a referee whose recommendation was reversed cannot referee that result again.
- Two different referees, counted by ORCID iD. Each must have been a referee of the problem's field when they signed. Revoking a referee later does not undo their recommendation. Problems cannot change field yet; when they can, only recommendations made by referees of the new field will count.
- Signed like every decision: staged, checked, signed in a batch with one ORCID sign-in, checked again inside the signature (signing). A referee whose appointment is revoked before the signature completes signs nothing.
Who may not referee a result
| Excluded | Example |
|---|---|
| Its author, by ORCID iD or by account. | A referee who posted the solution cannot recommend it. |
| The author of another live result on the same problem. | A referee with a progress result on the problem cannot referee a competing solution while their own result is live. |
| Anyone who signed the editor's decision on it (accept, reject or not accepted at screening), including one later superseded. | An editor who is also a referee of the field and accepted the solution cannot also recommend it. |
| A referee who already gave a recommendation on it, or whose recommendation on it was reversed. | A referee who signed "do not recommend" cannot later recommend. |
| The chief editor. | The chief editor never acts as a referee, even with a referee appointment, so that the chief editor stays free to decide disputes about recommendations. |
Citing someone's result in builds_on does not exclude them; the result's page in the queue shows whom it cites, and says so when it cites the viewer.
During PubPhys's launch, results, contests and comments open in a field once it has one editor (the chief editor counts). With one editor, a decision cannot yet be contested or appealed to anyone else: contests and appeals are published and wait for a second editor, with no deadline. Referees are appointed as the site grows; a solution is verified only after two referees recommend it.
A known gap. A field opens with 2 referees, but the count includes referees who are also the field's editors. An editor-referee who signed the acceptance of a result, or a referee with a live result of their own on the problem, cannot referee it, so fewer than two referees may be left for that result. It then waits, without a deadline, until the chief editor appoints another referee; the administrator is told by email of every accepted result with fewer than two referees who have recommended on it or still may. Counting only referees who are not the field's editors is an open question for the owner.
Who is excluded from what
Every rule is checked when a decision is staged and again inside the signature; removal and withholding are checked when the editor gives the instruction. Their own problems and results are left out of an editor's queue (a referee's queue lists them, with the reason); a result they are excluded from for another reason is listed, and its page gives the reason and offers no form for it. Nobody withholds the reasons of a decision they signed or instructed. PubPhys's own reviews of a seed problem's or a topic's revision have no human author: any editor of the field may withhold their reasons, from the record's page, except whoever wrote, signed or approved (in a seed batch) that revision; the audit log records the field at that moment. A removal record gives only its category, so there is nothing to withhold; any other text that cannot be traced to a problem or a result is refused, and the administrator handles it.
| Who | Decide on a problem | Decide on a result | Recommend a result | Decide a contest of it (triage, outcome, credit, reversal) | Remove it or a contest, or withhold a text on it |
|---|---|---|---|---|---|
| Its author, by ORCID iD or by account | No | No | No | No | No |
| The author of another live result on the same problem | Yes | No | No | No | Not the result, a contest of it, nor a decision or withdrawal on it; nor the problem or any later revision of it |
| The holder of a sealed result on the problem, by ORCID iD (unrevealed, lapsed or abandoned) | Not a later revision while the seal is unrevealed or lapsed, nor moving the problem to another topic | No | No | No | Not the problem, a later revision, a result, a contest or a text on them |
| Whoever signed the editor's decision on the result | — | — (one decision stands) | No | No, even after it is reversed | Yes |
| Whoever signed a decision a contest is against (a recommendation, a credit) | — | — | — | Not that contest | Yes |
| Whoever signed an outcome of a contest of the result (uphold, dismissal, credit, reversal) | — | Yes | — | No | Yes |
| The editor who triaged a contest of the result | — | Yes | — | Not that contest's outcome; other contests, yes | Yes |
| The filer of a contest | — | Yes | — | Not that contest | Not that contest, nor a text in it |
| A referee who already recommended it, or whose recommendation was reversed | — | — | No | Not a contest of the result or of their recommendation | Yes, as an editor of the field |
| The editor whose acceptance of it was reversed | — | No | No | Not the contest against it | Yes |
| Whoever decided an appeal of a decision on a contest | — | — | — | Not that contest | Yes |
| Whoever signed a decision an appeal granted, or signed the grant | Not the fresh decision | Not the fresh decision | — | Not the fresh triage or outcome | Yes |
| The chief editor | Yes | Yes | No | Yes | Yes |
Someone the result cites in builds_on | Yes | Yes | Yes | Yes | Yes |
A sealed result stays private, so its holder's exclusion is silent: the holder alone sees the reason, and nothing else changes. It raises no "no eligible editor" flag, adds no "excluded by interest" marker, and never refuses an appointment. When only seals keep every editor from an item, the administrator sees the item, by its link alone, in a private list ("No editor can decide this item; lend one") and asks the chief editor to lend an editor. What stays inferable is exactly this:
- an editor who does not act on an item, which looks the same as an editor who chooses not to;
- the administrator's list of privately blocked items, which shows that some editor of an item is privately excluded, but not who or why;
- an item that stays undecided longer than usual.
Nothing else differs from the absence of a seal (sealing).
Removing a contest that is undecided, or that has an appeal of one of its decisions waiting, counts as deciding it: the column for a contest applies (an open contest's triage editor included), and the signer of the appealed decision is excluded too. Deciding an appeal: no one who decided anything in the case, and a removal or withholding in the case counts as deciding (appeals).
A topic's revision, or an appeal of a decision on it, is not decided by whoever proposed the topic or signed that revision or an undecided one it builds on (by ORCID iD or by account), nor by the author of a problem waiting under the topic (proposing a topic).
Examples: an editor whose progress result on a problem is live cannot accept a competing solution there, nor remove it or withhold the text of its withdrawal, but can still accept or reject the problem itself. The author of a result cannot withhold the reasons of the editor's decision on their own result, even if they are also an editor of the field. "Yes" assumes the role the decision needs: an editor of the field (or the chief editor) for decisions, removal and withholding; a referee of the field for recommendations. Each contest of a result is decided by editors with no stake in it: every editor who decided an outcome of one contest is out of the others, so a result with many contests needs more editors (the chief editor counts). Removing an undecided contest closes it, so only an editor who could decide it may remove it (not the signer of the decision it contests); the same holds while an appeal of a decision on it waits, and the signer of the appealed decision may not remove it either. An appeal is decided by an editor who decided nothing in the case (appeals).
Reasons
Every signed decision carries the editor's or referee's reasons, at most 4,000 characters of plain text. They are signed, so they cannot be changed afterwards, and they are public whenever the item is public. Editors must not put personal data in them (for "not accepted at screening" they are optional). If they contain copyright material, personal data, abuse or spam anyway, an editor of the field who is not excluded from the item (above) can withhold them for good; the decision stays in force, and a removal record written by PubPhys on that editor's instruction says so (removal and withheld texts).
What the records show
A decision is a review record whose author is the editor's or referee's ORCID iD and whose target is the fingerprint
of the problem's signed revision or of the result's claim record; its content is the decision and the reasons. Anyone can check it with the record's bundle, without
trusting PubPhys (the protocol). Before phase 4 these decisions were
written by PubPhys on an editor's instruction; those older records name no editor, and PubPhys's audit log keeps who gave the
instruction.