Evidence guide · version lineage

One work, two versions: how to cite an evolving preprint

A registry-backed guide to what changed in the public record for The Reinforcement Learning Contour—and what the record does not authorize anyone to claim.

Status: Version 2 preprint · Conceptual research · Not peer-reviewed · One evolving work · No citation-count claim

The public-record change

Crossref and Europe PMC now expose two DOI-version records for the same title and sole author. Version 1 was posted on 4 August 2026; Version 2 was posted on 10 August 2026. The canonical research catalogue therefore points to Version 2 while preserving Version 1 in the work's version history.

RecordPostedEvidence-backed interpretation
Version 1 DOI4 August 2026Earlier public version of the same conceptual preprint.
Version 2 DOI10 August 2026Latest registry-observed public version; still a preprint and not peer-reviewed.

Why this is one work, not two publications

The title and author are unchanged, and the publisher-deposited identifiers share a manuscript lineage. A version-specific DOI identifies the exact public state being cited; it does not turn a revision into a separate research work. Counting both records as separate publications would overstate the corpus.

What the registries establish—and what they do not

Established

Identity and status

The checked registry metadata supports the title, Emanuel Shirbint as sole author, the two posting dates, the two version-specific DOIs, and classification as a preprint.

Not established

Substantive change or validation

Registry metadata alone does not prove which arguments changed, that Version 2 is empirically stronger, that peer review occurred, or that any framework was validated in production.

A safe citation rule

  1. Cite the exact version read, using its version-specific DOI.
  2. Label it as a preprint and preserve the “not peer-reviewed” status.
  3. Use Version 2 as the current catalogue target while retaining Version 1 as provenance.
  4. Do not publish citation counts until registry coverage and counting rules are reconciled.

Claim boundaries

This update changes a preprint-version record only. It does not change or substantiate peer-reviewed work, conference work, production systems, employment history, or advisory work. Those remain separate evidence classes with separate sources. It also does not claim that the manuscript is a deployed system, an evaluated production intervention, or a peer-reviewed result.

Primary records

Evidence checked 14 August 2026 (Asia/Jerusalem). This page documents publication status and version lineage; it does not infer unverified manuscript changes.