Latest Posts by @cigitalgem

NEW BIML Bibliography entryhttps://www.nature.com/articles/s41586-025-08661-4Mert Yuksekgonul et alNew computational paradigm of semantics of language. Auomatic differentiation by analogy. A very naive approach to feedback loops ignoring all kinds of known problems including: looping, non-convergence, and local minima. #MLsec #ML https://berryvilleiml.com/references/ ...
BIML has read this one three timeshttps://berryvilleiml.com/bibliography/Raji 2021 — Evaluation is HardRaji, Inioluwa Deborah, Emily M. Bender, Amandalynne Paullada, Emily Denton, Alex HannaAI and the Everything in the Whole Wide World Benchmark” arXiv preprint arXiv:2111.15366 (2021).Turns out that SOTA tests are not good. Evaluation of actual skills can’t be done ...