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 2112.07899cs.IRcs.CL
收藏

大型双编码器是可推广的检索器

Large Dual Encoders Are Generalizable Retrievers

Google Research(谷歌研究院)

Jianmo Ni, Chen Qu, Jing Lu, Zhuyun Dai, Gustavo Hernández Ábrego, Ji Ma, Vincent Y. Zhao, Yi Luan, Keith B. Hall, Ming-Wei Chang, Yinfei Yang

摘要:

已经表明,在一个域上训练的双编码器通常不能推广到其他域的检索任务。一个普遍的看法是,双编码器的瓶颈层,其中最终得分只是一个查询向量和通道向量之间的点积,是太有限,使双编码器的有效检索模型外的领域泛化。在本文中,我们挑战这种信念,规模的双重编码器模型{ \em的大小,同时保持瓶颈嵌入大小固定。}令人惊讶的是,通过多阶段训练,扩大模型大小可以显着改善各种检索任务,特别是对于领域泛化。实验结果表明,我们的双编码器,\textbf{G}可推广的\textbf{T} 5为基础的密集\textbf{R}检索器(GTR),优于%ColBERT~ \cite{khattab2020colbert}和现有的稀疏和密集检索器的BEIR数据集~ \cite{thakur2021beir}显着。最令人惊讶的是,我们的消融研究发现,GTR是非常有效的数据,因为它只需要10%的MS Marco监督数据,以实现最佳的域外性能。所有的GTR型号都在https://tfhub.dev/google/collections/gtr/1上发布。

英文摘要:

It has been shown that dual encoders trained on one domain often fail to generalize to other domains for retrieval tasks. One widespread belief is that the bottleneck layer of a dual encoder, where the final score is simply a dot-product between a query vector and a passage vector, is too limited to make dual encoders an effective retrieval model for out-of-domain generalization. In this paper, we challenge this belief by scaling up the size of the dual encoder model {\em while keeping the bottleneck embedding size fixed.} With multi-stage training, surprisingly, scaling up the model size brings significant improvement on a variety of retrieval tasks, especially for out-of-domain generalization. Experimental results show that our dual encoders, \textbf{G}eneralizable \textbf{T}5-based dense \textbf{R}etrievers (GTR), outperform %ColBERT~\cite{khattab2020colbert} and existing sparse and dense retrievers on the BEIR dataset~\cite{thakur2021beir} significantly. Most surprisingly, our ablation study finds that GTR is very data efficient, as it only needs 10\% of MS Marco supervised data to achieve the best out-of-domain performance. All the GTR models are released at https://tfhub.dev/google/collections/gtr/1.

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。

↑