arXiv - CSCL: "Why Can GPT Learn In-Context? Language Models Imp…" - Qoto Mastodon

arXiv - CSCL @arxiv_cscl@qoto.org

Why Can GPT Learn In-Context? Language Models Implicitly Perform Gradient Descent as Meta-Optimizers. (arXiv:2212.10559v3 [cs.CL] UPDATED)

http://arxiv.org/abs/2212.10559 #arXiv #NLProc

May 16, 2023, 03:07 · · arxiv-cscl · · ·

Sign in to participate in the conversation