The Ghost Couple: Correlated LLM Name Priors and Their Haunting of the Web and Academic Publishing

2026-06-01 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, AI Content Integrity · Depth: Expert, medium

Summary

A study by Michał Brzozowski and Neo Christopher Chung reveals that large language models (LLMs) generate not just high-probability individual names, but correlated character ensembles, such as "Elena Vasquez + Marcus Chen + Amara Okafor" for Claude, "Aris Thorne + Lena Petrova" for Gemini, and "Elara Voss" for GPT. These name priors are model-family and version-specific, showing consistent co-occurrence rates across independent generations. The research notes that these priors are actively suppressed at model release boundaries, leaving identifiable behavioral fingerprints. A significant downstream consequence is documented on Zenodo, a CERN-operated repository, where 1,655 "ghost-authored" records were found, claiming nonexistent journals with fabricated publication dates. DataCite timestamps confirm deliberate backdating, with 991 records registered in a single month, all carrying real DOIs. Additionally, ghost names appear on ResearchGate, forming synthetic research groups, with publication dates providing a temporal proxy for model deployment.

Key takeaway

For research scientists and academic publishers evaluating scholarly integrity, this research highlights a critical vulnerability: LLM-generated ghost authors are infiltrating repositories like Zenodo with fabricated publications. You should implement robust verification processes for new submissions, cross-referencing author names and journal details against established databases. Be wary of records with suspicious publication dates or correlated author ensembles, as these indicate potential AI-generated content that could undermine research credibility and data integrity.

Key insights

LLMs generate correlated fictional character ensembles, not just individual names, leading to widespread ghost authorship in academic repositories.

Principles

• LLM name generation exhibits model-family and version-specific correlations.
• Model release boundaries suppress prior generation behaviors.
• Fictional LLM-generated entities can propagate into real scholarly infrastructure.

Method

The study identified correlated name priors by analyzing independent LLM generations and traced their downstream impact by scanning Zenodo and ResearchGate for ghost-authored records and fabricated publication metadata.

In practice

• Monitor academic repositories for correlated fictional author names.
• Use DataCite timestamps to detect backdated or fabricated records.
• Analyze publication dates of ghost-authored content for LLM deployment proxies.

Topics

• Large Language Models
• Ghost Authorship
• Academic Integrity
• Zenodo Repository
• DataCite DOIs
• AI Misinformation

Code references

• tinierZhao/Academic-Industrial-associations

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Scientist, Research Scientist, AI Ethicist

Related on AIssential

• Spectral Audit of In-Context Operator Networks — Prediction accuracy alone is insufficient for neural operators; local operator fidelity requires spectral auditing.
• Bridging the Sim-to-Real Gap in Semiconductor Visual Program Synthesis via Input Binarization — Input binarization effectively bridges the sim-to-real gap for VLMs in semiconductor visual program synthesis by abstracting texture.
• Topological texture analysis of microscopy images of dynamic casein gelation and its relation to rheological properties — The toolbox combines TDA, DBC, MFP, and LBP to quantify dynamic microstructural changes in gelation, correlating with rheological properties.
• Conditional Monge Gap enables generalizable single-cell perturbation modelling — CMonge is a conditional optimal transport framework that accurately predicts single-cell perturbation responses and generalizes to unseen conditions.
• Learning the coupled dynamics of global climate modes — UniCM unifies global climate mode forecasting by learning coupled ocean-atmosphere dynamics and inter-mode interactions, revealing emergent predictability.
• Science Earth: Towards A Planet-Scale Operating System for AI-Native Scientific Discovery — A planet-scale runtime enables AI capabilities to self-organize for distributed, self-correcting scientific discovery.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.