Matrix Orthogonalization 도입으로 mLSTM Noisy AR 성능 최대 45.4%p 향상
Matrix Orthogonalization Improves Memory in Recurrent Models
Matrix Orthogonalization Improves Memory in Recurrent Models
How RNNs Work — Remembering Previous States in Sequential Data
Attention Mechanisms: Stop Compressing, Start Looking Back
"Attention Is All You Need" Paper tahun 2017 yang mengubah dunia kecerdasan buatan, dijelaskan tanpa perlu latar belakang teknis.
Introducing RWKV - An RNN with the advantages of a transformer