arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2202.07856cs.CLcs.LG
收藏

长距离Transformer的NLP任务有效性

The NLP Task Effectiveness of Long-Range Transformers

Johns Hopkins University(约翰斯·霍普金斯大学)

Guanghui Qin, Yukun Feng, Benjamin Van Durme

AI总结:

针对Transformer难处理长序列、长距离变体有效性研究不足的问题,该研究在多任务多数据集上基准测试7种长距离Transformer,分离预训练等干扰并分析注意力行为,明确了其优势与未被发现的缺陷。

摘要:

Transformer模型因O(N^2)的时间与空间复杂度,难以扩展至长序列场景。这催生了Longformer、Performer等旨在降低计算复杂度的Transformer变体。这类模型虽在理论上效率更高,但在真实NLP任务上的有效性尚未得到充分研究。我们在5项高难度NLP任务、7个数据集上对7种Transformer模型变体开展基准测试,设计实验分离预训练与超参数设置的影响,聚焦其长距离注意力能力。此外,我们提出多种研究注意力行为的方法,以揭示指标分数之外的模型细节。研究发现,长距离Transformer的改进注意力在内容选择、查询引导解码方面具备优势,但也存在此前未被认识到的缺陷,例如对远距离 token 注意力不足、累积近似误差。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑