The NLP Task Effectiveness of Long-Range Transformers

TL;DR

This paper benchmarks various long-range Transformer models on challenging NLP tasks, revealing their strengths in content selection and decoding, but also uncovering limitations like poor distant token attention and approximation errors.

Contribution

It provides a comprehensive empirical evaluation of 7 Transformer variants on multiple NLP tasks, analyzing their attention behaviors and identifying both advantages and drawbacks.

Findings

Long-range Transformers excel in content selection and query-guided decoding.

They exhibit insufficient attention to distant tokens.

Approximation errors accumulate, affecting performance.

Abstract

Transformer models cannot easily scale to long sequences due to their O(N^2) time and space complexity. This has led to Transformer variants seeking to lower computational complexity, such as Longformer and Performer. While such models have theoretically greater efficiency, their effectiveness on real NLP tasks has not been well studied. We benchmark 7 variants of Transformer models on 5 difficult NLP tasks and 7 datasets. We design experiments to isolate the effect of pretraining and hyperparameter settings, to focus on their capacity for long-range attention. Moreover, we present various methods to investigate attention behaviors to illuminate model details beyond metric scores. We find that the modified attention in long-range transformers has advantages on content selection and query-guided decoding, but they come with previously unrecognized drawbacks such as insufficient attention to distant tokens and accumulated approximation error.