# Scoring rubric

Seven dimensions, each scored from 1 to 5 against written anchors. Levels 2 and 4 are for manuscripts that sit between two anchors. A score is a judgment, not a measurement: it is only as good as the findings it is tied to, so each score comes with a reason that names them.

# Dimensions and weights

Total out of 100 = sum over dimensions of weight x (level - 1) / 4.

# Anchors

# 1. Contribution

• 1 The question is already answered in work the paper cites or should cite; nothing new is identified.
• 3 A clear increment (new data, setting or method applied to a known question); the difference from the closest papers is stated but its importance is not argued.
• 5 The paper changes what a reader in the field believes or can do; the difference from the closest work is stated precisely and the claim of novelty is backed by a documented comparison.

# 2. Method and identification

• 1 The design cannot answer the question asked (the comparison does not isolate what is claimed), or key steps are not described.
• 3 The design is appropriate and described; the main assumption is stated, but a plausible alternative explanation is not addressed, or inference ignores a feature of the data.
• 5 The design is well matched to the question, assumptions are stated and tested where testable, threats are named and handled, and inference fits the data structure.

# 3. Evidence and robustness

• 1 Main claims are not supported by the exhibits shown, or results contradict the text.
• 3 Main results support the central claim; robustness checks exist but leave out an obvious one, magnitudes are not interpreted, or secondary claims lack support.
• 5 Every claim in the abstract and conclusion is carried by an exhibit; checks that could have failed are reported, including unfavourable ones; magnitudes are interpreted.

# 4. Exposition and structure

• 1 The argument cannot be followed without rereading; sections are missing or misplaced; unfinished text remains.
• 3 Readable and complete; some sections do two jobs, the introduction does not state the results, or important material sits in the appendix.
• 5 A reader knows the question, the approach and the result after the introduction; each section does one job; nothing essential is in the appendix.

# 5. Exhibits

• 1 Tables or figures cannot be read alone, lack units or uncertainty, or disagree with the text.
• 3 Exhibits are correct and referenced; several lack complete notes, sample sizes or a legend, or are more numerous than needed.
• 5 Every exhibit is self-contained (title, units, uncertainty, N, notes), is called out, and matches the text number for number.

# 6. Integrity of citations and numbers

• 1 Citations do not resolve or do not support what is attributed to them; numbers disagree between abstract, text and tables.
• 3 References resolve and are complete; a few numbers or attributions could not be matched, with no pattern.
• 5 Every checked citation supports its sentence, every number in the abstract and conclusion is found in the results with the same value, and no internal contradiction was found.

# 7. Replicability

• 1 Data source, sample construction or estimation cannot be reconstructed from the paper; no code or data statement.
• 3 Data and method are described well enough to attempt a replication; code or data are promised or partly available; versions and seeds are missing.
• 5 Data (or a documented access route), code, environment and seeds are available and map to each exhibit; restrictions are explained.

# Caps

A total can hide a fatal flaw. Two caps apply after the sum:

• If Method and identification is 1, the total is capped at 40.
• If Integrity is 1 (fabricated or unsupported references, numbers that cannot be reconciled), the total is capped at 40 and the editor is told separately.

State when a cap was applied.

# Reading the total

The total orders manuscripts roughly; it does not decide. Indicative bands:

The recommendation follows from the major comments and whether they can be resolved, and it can depart from the band: say why when it does. Different venues set different bars for contribution; adjust the reading of dimension 1, not the anchors.

# Reporting

Dimension                         Level  Weight  Points  Reason (findings)
1 Contribution                      3      20     10.0   E-01, E-02
2 Method and identification         2      25      6.25  R-01 (major), R-04
3 Evidence and robustness           3      20     10.0   R-02, R-05
4 Exposition and structure          4      10      7.5   E-05
5 Exhibits                          3      10      5.0   T-02, T-03, A-07
6 Integrity                         4      10      7.5   I-01
7 Replicability                     2       5      1.25  P-01, P-02
Total                                             47.5   no cap applied

For a revision, add a column with the previous level and name the findings that were resolved. Do not average the scores of several reviewers without reading why they differ: a large gap on one dimension is itself information for the editor.

The hygiene score printed by the audit script is unrelated to this rubric. It counts mechanical defects and must not be reported as a quality score.