claim-auditor

To verify manuscript claims, the claim-auditor extracts auditable claim units from text and matches them against repository artifacts, experiment outputs, and verified citations. It then assigns support labels to indicate whether each claim is fully supported, partially supported, or contradicted.

Auditing scientific writing for overclaiming involves breaking down text sentence-by-sentence to classify claim types like performance, novelty, and robustness. The tool checks each claim against available evidence sources and recommends conservative rewrites to prevent unsupported statements.

Yes, you can check citation context and evidence support automatically by extracting auditable claim units from contributions or related work sections. The system locates the strongest matching evidence across literature sources and project context to verify if citations correctly support the claims.

To identify unsupported statements in a research report, the system assigns nuanced support labels including unsupported, contradicted, or needs manual check. It evaluates abstracts, results summaries, and conclusions against repository evidence to pinpoint exactly where claims lack necessary backing.

Evidence verification works for rebuttals and reviewer-sensitive statements by applying special handling to novelty, performance, robustness, and generalization claims. It matches these specific statements to appropriate evidence hierarchy levels and suggests required follow-up steps if support is missing.

Automated claim auditing for research integrity does not invent support for claims. When evidence is missing or ambiguous, it assigns labels like needs manual check or preliminary, and recommends conservative rewrites or next-step skills rather than fabricating verification.