Aditya Varre

Incremental Learning in Transformers for In-Context Associative Recall

Learning In-context $n$-grams with Transformers: Sub-$n$-grams Are Near-Stationary Points