Alicia Curth and colleagues presented work arguing that double descent is an illusion caused by using parameter count as an inadequate measure of model complexity.
Sources
- Large language models can do jaw-dropping things. But nobody ... (technologyreview.com)
- https://arxiv.org/abs/2310.18988 (arxiv.org)
- https://neurips.cc/virtual/2023/poster/71830 (neurips.cc)
- https://proceedings.neurips.cc//paper_files/paper/2023/hash/aec5e2847c5ae90f939ab786774856cc-Abstract-Conference.html (proceedings.neurips.cc)
More in Science
- The reduction in amyloid deposits from estrogen skin patch use was most apparent among women carrying the APOE e4 genotype, a genetic variant associated with greater risk of late-onset Alzheimer's disease.
- Traces of SARS-CoV-2 were detected in open-sea water samples collected at different sites during the Sea Care project.
- Researchers used fossil foraminifera species in sediment cores as indicators of past ocean temperatures, since some species prefer warm water while others favor cooler conditions.
- The NIA Interventions Testing Program found that acarbose significantly extended lifespan in both sexes of UM-HET3 mice, but the effect was much greater in males than in females (Harrison et al., 2014).
- In the Da Qing Diabetes Prevention Study, participants receiving lifestyle intervention had an average increase in life expectancy of 1.44 years (95% CI 0.20–2.68; p=0.023).
- Researchers described the period 5-36 months past menopause as the 'critical window' of rapid estrogen depletion.