readme

README.md

@@ -317,6 +317,8 @@ model = TransformerWrapper(
 
 Update: MetaAI researchers <a href="https://arxiv.org/abs/2309.16588">have found</a> that adding memory tokens (they call them register tokens), alleviates outliers (which is suspected now to be a pathology of attention networks unable to <a href="https://arxiv.org/abs/2306.12929">attend to nothing</a>).
 
+Update 2: a hybrid architecture out of Nvidia named <a href="https://openreview.net/forum?id=A1ztozypga">Hymba</a> used memory tokens successfully in the autoregressive case, termed meta tokens in their paper
+
 ### Transformers Without Tears
 
 <img src="./images/scalenorm.png"></img>