arXiv - CSCL: "Never Train from Scratch: Fair Comparison of Long…" - Qoto Mastodon

arXiv - CSCL @arxiv_cscl@qoto.org

Never Train from Scratch: Fair Comparison of Long-Sequence Models Requires Data-Driven Priors. (arXiv:2310.02980v2 [cs.LG] UPDATED)

http://arxiv.org/abs/2310.02980 #arXiv #NLProc

Dec 05, 2023, 03:20 · · arxiv-cscl · · ·

Sign in to participate in the conversation