Table Duplication

Copy-pasted rows and columns are one of the clearest signatures of a fabricated table. ReviewerZero extracts every table in your manuscript and checks it against every other table in the paper, flagging duplicated content that would be nearly impossible to spot by eye.

What We Detect

Duplicated columns - Two "independent" groups reporting identical values down to the last decimal
Duplicated rows - The same outcome row reused across tables
Duplicated blocks - Larger regions of one table copied into another
Partial duplication - Overlaps that cover only part of a table, where the rest was edited

How It Works

Table Extraction

Every table is extracted from the manuscript during document parsing, preserving its cell structure and numeric content.

Overlap Detection

Deterministic matching algorithms compare the numeric grid of every table against every other table in the manuscript, so a match is an exact computational finding — not a judgment call.

Legitimacy Assessment

Not every repeated value is a problem: baseline characteristics are often repeated by design, and comparison tables can legitimately share a control group. A multimodal AI pass reviews the actual page images of each candidate match and filters out overlaps with a legitimate explanation. Only non-legitimate duplication is flagged.

Reviewing Results

Findings appear in the Tables tab of your report:

Each finding shows the matched tables and the overlapping region
Click through to jump to the tables' locations in the manuscript
Dismiss false positives with one click, and restore dismissed findings anytime
Table duplication findings are included in exported reports and in the API response

Severity

High-severity duplication across tables is treated as a red finding — it appears in the "Needs review" check-visibility level alongside the other decision-relevant issues.

Limitations

Comparison happens within a single manuscript; tables are not checked against other papers
Tables that render as images rather than text may not be fully extractable
Legitimate-repetition filtering is conservative; when in doubt, a finding is surfaced for human review

Learn More

Read our blog post on catching partially duplicated tables: How We Catch Partially Duplicated Tables.

Statistical Checks - Verification of reported statistics, including table values
Image Analysis - Figure duplication detection
Export & Reports - Download findings in PDF, Word, and PowerPoint

Table Duplication

On this page