arXiv - CSCL: "Sequence-to-Sequence Pre-training with Unified Mo…" - Qoto Mastodon

arXiv - CSCL @arxiv_cscl@qoto.org

Sequence-to-Sequence Pre-training with Unified Modality Masking for Visual Document Understanding. (arXiv:2305.10448v1 [cs.CL])

http://arxiv.org/abs/2305.10448 #arXiv #NLProc

May 19, 2023, 03:07 · · arxiv-cscl · · ·

Sign in to participate in the conversation