bibguard

Detect hallucinated and broken citations in academic papers.

TypeScript implementation — works in Node.js (>=18) and modern browsers. Zero dependencies.

npx bibguard paper.bib

Install

# CLI (no install needed)
npx bibguard paper.bib

# As a library
npm install bibguard

CLI Usage

npx bibguard references.bib
npx bibguard references.bib --json --out report.json

Library Usage

import { parseBib, verifyAll } from "bibguard";

// Parse .bib text
const entries = parseBib(bibFileContent);

// Verify all entries (with progress callback)
const results = await verifyAll(entries, (i, total, key, status) => {
  console.log(`[${i}/${total}] ${key}: ${status}`);
});

// Check results
for (const r of results) {
  if (r.overall === "FAIL") {
    console.log(`${r.key}: ${r.checks.filter(c => c.status === "FAIL").map(c => c.field).join(", ")}`);
  }
}

Browser Usage

All 5 API sources support CORS — bibguard works directly in the browser without a proxy:

<script type="module">
  import { parseBib, verifyEntry } from "https://esm.sh/bibguard";

  const bib = `@article{test, title={Attention Is All You Need}, author={Vaswani}, year={2017}}`;
  const entries = parseBib(bib);
  const result = await verifyEntry(entries[0]);
  console.log(result.overall, result.sourcesHit);
</script>

This makes it possible to build:

• Browser extensions for reviewing papers
• Overleaf plugins for real-time citation checking
• Zotero/Mendeley integrations
• VS Code extensions for .bib file validation

API Sources

| Source | Lookup | CORS | Coverage | |--------|--------|------|----------| | arXiv | ID resolution | Yes | CS, Physics, Math | | Crossref | DOI resolution | Yes | 150M+ records | | DBLP | Title search | Yes | CS papers | | Semantic Scholar | Title search | Yes | 200M+ papers | | OpenAlex | Title search | Yes | 250M+ works |

All queries respect rate limits. No API keys required.

Features

• Phantom ID detection: DOI/arXiv that looks valid but doesn't resolve
• Kill-shot logic: Phantom IDs can never be overridden by search results
• Author guard: Stricter matching when phantom IDs are detected
• Auto-fix suggestions: Missing DOIs collected from API results
• Zero dependencies: Only uses fetch (built into Node 18+ and browsers)

Also available in Python

pip install bibguard

See bibguard (Python) for the full Python version with TeX cross-audit, duplicate detection, Markdown reports, and .bib auto-fix.

Related

• IntegriRef — Full L0-L4 verification stack (semantic NLI, citation graph, Bayesian risk)

License

Apache License 2.0

bibguard (TypeScript 版)

TypeScript 实现的学术引用验证工具。支持 Node.js 和浏览器环境，零依赖。

npx bibguard paper.bib

所有 5 个 API 源均支持 CORS，可直接在浏览器中运行——适合构建审稿人插件、Overleaf 扩展、Zotero 集成等。

Python 版请使用 pip install bibguard。