AI Hallucinations Plague Prestigious NeurIPS Conference Papers
AI
GPTZero
Hallucinations
NeurIPS
Startups
Artificial Intelligence
Research
8
Verification Gap
Media Hype
6/10
Real Impact
8/10
What is the Viqus Verdict?
We evaluate each news story based on its real impact versus its media hype to offer a clear and objective perspective.
AI Analysis:
While the finding itself isn’t dramatically novel, the scale of the issue – 100 fabricated citations from top researchers – is generating substantial concern and driving a critical conversation about the verification of AI-generated content, representing a significant challenge to the integrity of academic research.
Article Summary
GPTZero’s analysis of papers accepted by the Conference on Neural Information Processing Systems (NeurIPS) revealed a concerning number of hallucinated citations. The startup identified 100 instances of fabricated citations across 51 papers, confirming them as entirely fake. This finding highlights a potential issue within the increasingly reliance on Large Language Models (LLMs) for tasks like citation generation, particularly among the world’s foremost AI researchers. While NeurIPS argues that even a small percentage of inaccurate references doesn't invalidate the research’s core findings, the sheer volume of citations generated by LLMs presents a significant challenge to verification. The discovery underscores the strain on conference review pipelines and raises questions about the rigorousness of academic publishing in the age of AI. The situation is further complicated by the importance of citations as a metric of a researcher’s influence and impact within the field. The issue has even been previously discussed, as noted by a 2025 paper, ‘The AI Conference Peer Review Crisis’Key Points
- 100 confirmed instances of fabricated citations were found across 51 papers submitted to NeurIPS, according to GPTZero.
- Despite NeurIPS' assertion that a small percentage of inaccurate references doesn’t invalidate the research, the large volume of AI-generated citations poses a significant verification challenge.
- The finding raises concerns about the accuracy of AI-generated content within leading academic research, particularly given the reliance on LLMs and the importance of citations as a career metric.