Change the Product, Keep the Parameters: Associative Algebra Layers for Transformers
Algebraic feed-forward layers lift small Transformer throughput 6–8% but cut downstream scores.
2 stories, newest first · one card per story
Algebraic feed-forward layers lift small Transformer throughput 6–8% but cut downstream scores.
Algebraic Transformer layers raise small-model generation throughput 6.2-7.8% but lower downstream scores.
Today 1.6K readers · 2.5K views · 778 reads · all time 189K views