A survey identified data leakage in 22 papers across 17 research fields, collectively affecting 294 papers.
Sources
- Leakage and the reproducibility crisis in machine-learning-based science (cell.com)
- https://pmc.ncbi.nlm.nih.gov/articles/PMC10499856/ (pmc.ncbi.nlm.nih.gov)
- https://pubmed.ncbi.nlm.nih.gov/37720327/ (pubmed.ncbi.nlm.nih.gov)
- https://www.sciencedirect.com/science/article/pii/S2666389923001599 (sciencedirect.com)
- https://arxiv.org/pdf/2602.17854 (arxiv.org)
More in Science
- The neutrino beam used by ANNIE originates at Fermilab, where the Booster delivers 8 GeV protons to a beryllium target, producing secondary pions and kaons.
- Many scientific proposals have been made for different stages of the abiogenesis process.
- The 'Sapphire Canyon' rock sample was identified after peer review as containing potential biosignatures, clues that past life may have been present on Mars, though not proof of life.
- Under CDC's updated 2024 COVID-19 guidance, people may resume regular activities within 24 hours if they have been fever-free for at least 24 hours without medication and their symptoms are mild and improving.
- The NCEP document titled 'ENSO: Recent Evolution, Current Status and Predictions' is updated every Monday.
- A U.S. version of the affirmative-action replication study essentially reproduced the original result, but the Open Science Collaboration excluded this successful U.S. replication and included only the failed University of Amsterdam replication in its estimate of psychology's reproducibility.