NeurIPS started desk-rejecting papers over references that don't exist (strictcite.com)

The NeurIPS conference has started a new policy where papers containing fabricated or "hallucinated" citations may face automatic desk rejection. This shift comes as part of a broader trend among leading AI/ML conferences like ICLR, ICML, and ACL, which have also enacted similar measures. Specifically, NeurIPS 2026 is reportedly employing an automated reference checker that flags submissions with two or more unverified citations, taking the claim of "hallucinated references" seriously enough to influence acceptance rates significantly. Early indications suggest that this could have affected approximately 5% of submissions in previous cycles, emphasizing the growing importance of accurate bibliography in academic papers. This change is noteworthy for the AI/ML community as it represents a major step toward ensuring academic integrity and accountability. By leveraging automation combined with human review processes, conferences aim to filter out submissions based on the reliability of their references, despite acknowledging potential false positives in their systems. This move could reshape submission best practices, pushing authors to meticulously verify their citations before submitting, and potentially raising the overall quality of published research in the field.