Representation Collapse in Machine Translation Through the Lens of Angular Dispersion

TL;DR

This paper investigates the phenomenon of representation collapse in neural machine translation models, analyzing its dynamics and proposing angular dispersion regularization to mitigate it and enhance translation quality.

Contribution

It introduces an analysis of representation collapse in NMT models and demonstrates that angular dispersion regularization effectively mitigates collapse and improves translation performance.

Findings

Regularization based on angular dispersion reduces representation collapse.

Quantized models exhibit similar collapse behavior as full models.

Regularization benefits persist even after model quantization.

Abstract

Modern neural translation models based on the Transformer architecture are known for their high performance, particularly when trained on high-resource datasets. A standard next-token prediction training strategy, while widely adopted in practice, may lead to overlooked artifacts such as representation collapse. Previous works have shown that this problem is especially pronounced in the representation of the deeper Transformer layers, where it often fails to efficiently utilize the geometric space. Representation collapse is even more evident in end-to-end training of continuous-output neural machine translation, where the trivial solution would be to set all vectors to the same value. In this work, we analyze the dynamics of representation collapse at different levels of discrete and continuous NMT transformers throughout training. We incorporate an existing regularization method based on angular dispersion and demonstrate empirically that it not only mitigates collapse but also improves translation quality. Furthermore, we show that quantized models exhibit similar collapse behavior and that the benefits of regularization are preserved even after quantization.