arXiv - CSCL: "Research without Re-search: Maximal Update Parame…" - CleverLibre Social

arXiv - CSCL @arxiv_cscl@social.cleverlibre.org

Research without Re-search: Maximal Update Parametrization Yields Accurate Loss Prediction across Scales. (arXiv:2304.06875v3 [cs.CL] UPDATED)

http://arxiv.org/abs/2304.06875 #arXiv #NLProc

Sep 06, 2023, 03:17 · · arxiv-cscl · · ·

Sign in to participate in the conversation