Nekropolis: a cross-domain dataset of computer-science retractions and AI-governance lifecycle records (2018–2026)

Read full text
Authors: Anton Sokolov
Cited by: 1
References: 4

Abstract

Nekropolis is a curated public-record corpus of 14,923 artifacts (2018-2026) that were retracted, withdrawn, revoked, archived, or discontinued across computer-science research and adjacent digital-governance infrastructure, under one entry schema. The corpus is dominated by computer-science/AI-related journal retractions: the Retraction Watch slice is 90.5% of the release and is described throughout as an AI/ML/CS-related candidate slice, not an adjudicated AI-only set. Source families include journal retractions, preprint withdrawals, public retraction-flag metadata, trusted-service lifecycle changes, archived governance repositories, withdrawn benchmarks, closed European Digital Identity Wallet pilots, and one rescinded public model demo. Each record carries an artifact category, lifecycle dates, a source-stated reason where available, a conservative project-inferred cause label, and hashed source pointers. The corpus is descriptive; it is not an assessment of researchers, institutions, publishers, repositories, or service providers. The dataset can be reused for source-stratified study of negative governance evidence, temporal-awareness tests in retrieval systems, and evaluation of systems that extract evidence from public lifecycle records while preserving abstention when sources are silent. A versioned Zenodo package provides the release file, candidate-holdout file, schema, pipelines bundle, build notes, and manuscript source. The counts describe one curated release slice built under named gates and are not prevalence estimates.

Fields of Study & Subjects

Academic integrity and plagiarism Scientific Computing and Data Management Research Data Management Practices

Publication details

Journal / source
Data in Brief
Publisher
Elsevier BV
Volume
67
Pages
113041–113041
Publication date
27 June 2026
Language
EN