English

LLMs Have Made Failure Worth Publishing

Digital Libraries 2026-04-09 v1

Abstract

Scientific publishing systematically filters out negative results. We argue that this long-standing asymmetry has become an urgent problem in the era of large language models, which inherit the positive bias of the literature they are trained on, face an impending shortage of high-quality training data, and are increasingly deployed as both research tools and peer reviewers. We analyze three ways in which LLMs have changed the value of failure data and show that the systematic absence of such data degrades their utility as research tools, training data consumers, and peer reviewers alike. We outline experimental protocols to validate these claims and discuss the structural conditions under which a failure-inclusive publishing culture could emerge.

Keywords

Cite

@article{arxiv.2604.06236,
  title  = {LLMs Have Made Failure Worth Publishing},
  author = {Sungmin Lee},
  journal= {arXiv preprint arXiv:2604.06236},
  year   = {2026}
}
R2 v1 2026-07-01T11:57:59.786Z