- Kevin Zhu, a Computer Science graduate (UC Berkeley), caused a stir by claiming authorship of 113 AI research papers in 2025 alone, with 89 accepted at the NeurIPS conference – one of the world’s largest AI events.
- Zhu, currently running the company Algoverse, charges 3,325 USD for each high school or university student participating in a “12-week research mentorship program,” with the goal of helping them publish at major conferences. He says he only “supervises, provides feedback on methodology, and edits drafts using AI tools or proofreaders.”
- Professor Hany Farid (UC Berkeley) condemned these works as a “disaster, just vibe coding” – meaning using AI to create software or articles without a scientific foundation. He warned that the phenomenon of fake research, idea plagiarism, and reliance on generative AI is discrediting the entire industry.
- The number of papers submitted to AI conferences is exploding: NeurIPS received 21,575 papers in 2025, double that of 2020; ICLR 2026 expects nearly 20,000 papers, a 70% increase compared to the previous year. Many critics say these papers are “low quality, suspected to be written by AI.”
- Reviewers like Jeffrey Walling (Virginia Tech) admit the NeurIPS review process is “too fast and shallow,” with many papers only skimmed by PhD students, leading to a lack of in-depth evaluation and correction before publication.
- Farid and Walling stated that the phenomenon of “academic hyper-productivity” is being wrongly glorified, causing students and faculty to race for publication rather than conducting genuine research.
📌 The AI academic world is facing an “AI slop” crisis – a large volume of worthless research born from competitive pressure and the abuse of generative AI. As tens of thousands of “virtual” papers flood prestigious conferences, trust in scientific quality is wavering. There are warnings that “academic hyper-productivity” is being wrongly glorified, causing students and faculty to race for publication rather than conducting genuine research.
