Preprint
Aug 2026
Kathleen Writes: Autoregressive Generation and Data Scaling Without Attention
Four architectural additions that did not help, a computed lexicon reaching 94% of a learned table's top-1 accuracy at one fifth of the parameters, and a computed lexicon reaching 94% of a learned table's top-1 accuracy at one fifth of the parameters.
George Fountzoulas
· 0 citations