Technical

Srinivasa Reddy Kandi: Fake Citations Discovered in NeurIPS Research Papers

January, 22, 2026-02:52

Share: Facebook | Twitter | Whatsapp | Linkedin | Visits: 37469 | :2821

---

Fake Citations Discovered in NeurIPS Research Papers:

In a twist of irony, fabricated citations have been uncovered in papers accepted by NeurIPS, one of the most respected conferences in artificial intelligence research. AI-detection startup GPTZero reviewed all 4,841 papers presented at the Conference on Neural Information Processing Systems, held last month in San Diego, and identified 100 hallucinated citations spread across 51 papers. The company confirmed that these references were fake, according to information shared with TechCrunch.

Acceptance at NeurIPS is considered a major professional milestone in the AI community, as the conference showcases work from some of the world’s leading researchers. Given this expertise, it might be expected that authors would rely on large language models only for routine tasks—such as formatting citations—rather than risk errors.

However, the findings come with important context. The number of confirmed hallucinated citations is statistically minimal when compared with the tens of thousands of total references across all accepted papers. In other words, the issue affects only a very small fraction of the overall citation pool.

It’s also worth noting that incorrect references do not necessarily undermine the validity of the research itself. NeurIPS emphasized to Fortune, which first reported GPTZero’s analysis, that even if around 1.1% of papers contain one or more inaccurate citations due to LLM usage, the scientific contributions of those papers remain intact.

Still, fabricated citations are not a trivial matter. NeurIPS positions itself as a venue for rigorous scholarly work in machine learning and AI, and all submissions undergo peer review by multiple reviewers who are instructed to watch for hallucinations. Citations also play a critical role in academic careers, serving as a measure of a researcher’s influence and impact. When AI-generated tools invent references, they risk diluting the value and integrity of that system.

Author: Kandi Srinivasa Reddy, Srinivasa Reddy Kandi, #KandiSrinivasaReddy, #SrinivasaReddyKandi