Hierarchical Autoregressive Transformers: Combining Byte- and Word-Level Processing for Robust, Adaptable Language Models.
Pit Neitemeier, Bjrn Deiseroth, Constantin Eichenberg, Lukas Balles
Browse the full ICLR paper archive.
Pit Neitemeier, Bjrn Deiseroth, Constantin Eichenberg, Lukas Balles
Browse the full ICLR paper archive.