HalluCitation Matters: Revealing the Impact of Hallucinated References with 300 Hallucinated Papers in ACL Conferences
This study addresses the problem of “HalluCitation”—the citation of non-existent references—in academic publications by conducting the first large-scale empirical analysis of all papers accepted to the three premier computational linguistics conferences (ACL, NAACL, and EMNLP) in 2024–2025. Leveraging bibliographic comparison, citation validation, and text mining techniques, the authors systematically verify the authenticity of cited references. Their analysis reveals that nearly 300 papers contain at least one HalluCitation, with over 100 such instances originating from the main conference and Findings of EMNLP 2025, indicating a rapidly growing prevalence of this issue in the field. This work not only introduces the term “HalluCitation” to quantify its impact but also provides the first evidence of its widespread occurrence in top-tier venues, offering critical insights for enhancing scholarly integrity.