Multi-Stream Transformers

TL;DR

This paper introduces a Multi-stream Transformer architecture that splits the encoder into multiple streams to preserve and explore alternative hypotheses, leading to improved performance in transformer models.

Contribution

The paper proposes a novel multi-stream encoder design that enhances hypothesis exploration and improves transformer performance, with added skip connections for further gains.

Findings

Splitting encoder into multiple streams improves performance.

Adding skip connections enhances the model further.

Multi-stream approach effectively preserves alternative hypotheses.

Abstract

Transformer-based encoder-decoder models produce a fused token-wise representation after every encoder layer. We investigate the effects of allowing the encoder to preserve and explore alternative hypotheses, combined at the end of the encoding process. To that end, we design and examine a $\textit{Multi-stream Transformer}$ architecture and find that splitting the Transformer encoder into multiple encoder streams and allowing the model to merge multiple representational hypotheses improves performance, with further improvement obtained by adding a skip connection between the first and the final encoder layer.