Science or Slop?: Benchmarking and Mitigating Scientific Slop in AI-Generated Papers
AI-generated content, often called AI slop, is increasingly common everywhere, particularly in academia. Slop in AI-generated scientific papers, however, has more complex patterns that cannot be easily detected by existing token-based AI detectors. Each part of such a paper looks plausible while the scientific reasoning that connects the parts breaks down, which can mislead how
Reading in Plain English mode — toggle below to switch
AI-generated content, often called AI slop, is increasingly common everywhere, particularly in academia. Slop in AI-generated scientific papers, however, has more complex patterns that cannot be easily detected by existing token-based AI detectors. Each part of such a paper looks plausible while the scientific reasoning that connects the parts breaks down, which can mislead how
This research highlights the critical and growing problem of AI-generated 'slop' in scientific papers, particularly focusing on the breakdown of scientific reasoning. For ML engineers, understanding these failure modes is crucial for developing more robust and reliable AI systems, especially those involved in content generation or scientific discovery.