Skip to main content

LLM plugin for verifying academic citations against Crossref

Project description

LLM Citation Verifier

An LLM plugin that verifies academic citations against the Crossref database to catch hallucinated references in AI-generated content.

The Problem

AI research tools sometimes hallucinate citations - generating plausible-looking DOIs for papers that don't exist.

The Solution

This plugin automatically verifies citations in real-time, flagging fake DOIs and validating real papers with full metadata.

Installation

llm install llm-citation-verifier

Usage Examples

Real-Time Citation Verification

# Verify citations as the LLM generates content
llm -T verify_citation "What's new in dye sensitized solar cells? Check all the references." --td

# Single breakthrough with verification
llm -T verify_citation "What's one recent breakthrough in cancer immunotherapy? Cite just one paper and verify it." --td

Quality Control for AI Content

# Review suspicious AI-generated content
llm -T verify_citation "This AI tool cited these papers: 10.1038/nature12373 and 10.1234/fake.doi.2024. Check if they're real." --td

# Batch verification
llm -T verify_citation "Verify these DOIs from an AI summary: 10.1038/nature12373, 10.1126/science.abc123, 10.1234/fake.journal.2024" --td

Research Integrity

# Audit AI research tools
llm -T verify_citation "Tell me about recent AI alignment breakthroughs. Verify any papers you cite." --td

# Fact-check literature reviews
llm -T verify_citation "What are the latest developments in quantum computing? Make sure all citations are real." --td

What It Does

  • Catches fake citations - Flags DOIs that don't exist in Crossref
  • Validates real papers - Returns title, authors, journal, year
  • Real-time verification - Works during content generation
  • Prevents hallucination - Stops fake references from appearing

Example Output

Tool call: verify_citation({'doi': '10.1038/s41591-023-02452-7'})
{
  "verified": false,
  "doi": "10.1038/s41591-023-02452-7", 
  "error": "DOI not found in Crossref database - likely hallucinated"
}

Tool call: verify_citation({'doi': '10.1016/j.cell.2023.02.029'})
{
  "verified": true,
  "doi": "10.1016/j.cell.2023.02.029",
  "title": "Discovery of phage determinants that confer sensitivity to bacterial immune systems",
  "authors": "Avigail Stokar-Avihail, Taya Fedorenko, Jens Hör, et al.",
  "journal": "Cell",
  "year": "2023"
}

Use Cases

  • Publishing workflows - Verify citations before publication
  • AI content review - Quality control for AI writing assistants
  • Research integrity - Audit AI-generated literature reviews
  • Fact-checking - Validate suspicious citation claims

Development

git clone https://github.com/your-org/llm-citation-verifier
cd llm-citation-verifier
uv sync
uv pip install -e .
uv run pytest

Pro Tips

  • Use --td flag to see verification calls in real-time
  • Chain limits prevent infinite verification loops
  • Works with any topic - just add "verify citations" to your prompt
  • Catches both completely fake DOIs and real DOIs used in wrong context

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

llm_citation_verifier-0.1.2.tar.gz (4.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

llm_citation_verifier-0.1.2-py3-none-any.whl (6.1 kB view details)

Uploaded Python 3

File details

Details for the file llm_citation_verifier-0.1.2.tar.gz.

File metadata

File hashes

Hashes for llm_citation_verifier-0.1.2.tar.gz
Algorithm Hash digest
SHA256 4b5d9bd1c8511976f3a52bdf4276d027ff317438fafdaab87141d55c2ffa06f8
MD5 a8b38c7eac8f95ea5d36619601fe5dfa
BLAKE2b-256 8370349e81b35388547ecb395035133df759487d009e0fee0c2c4677c355d801

See more details on using hashes here.

File details

Details for the file llm_citation_verifier-0.1.2-py3-none-any.whl.

File metadata

File hashes

Hashes for llm_citation_verifier-0.1.2-py3-none-any.whl
Algorithm Hash digest
SHA256 a867b18048e383ea95bb2d9dc530900674a46f9fe95f13758308e189bc7c092e
MD5 754b4a104cc33c215cc39bd2eb32037a
BLAKE2b-256 e01ef3c83c6943407384413b471e1c0fb319706880df5efb7288ebef7b7bc596

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page