Towards Dynamic Dense Retrieval with Routing Strategy

TL;DR

This paper introduces dynamic dense retrieval (DDR), a flexible and parameter-efficient method that adapts to new domains using prefix tuning and routing, outperforming traditional dense retrieval in zero-shot tasks.

Contribution

The paper proposes DDR, a novel dense retrieval framework combining prefix tuning and routing for efficient domain adaptation without retraining from scratch.

Findings

DDR surpasses traditional DR on six zero-shot tasks.

Uses only 2% of training parameters compared to full models.

Enables flexible and efficient domain adaptation in IR.

Abstract

The \textit{de facto} paradigm for applying dense retrieval (DR) to new tasks involves fine-tuning a pre-trained model for a specific task. However, this paradigm has two significant limitations: (1) It is difficult adapt the DR to a new domain if the training dataset is limited. (2) Old DR models are simply replaced by newer models that are trained from scratch when the former are no longer up to date. Especially for scenarios where the model needs to be updated frequently, this paradigm is prohibitively expensive. To address these challenges, we propose a novel dense retrieval approach, termed \textit{dynamic dense retrieval} (DDR). DDR uses \textit{prefix tuning} as a \textit{module} specialized for a specific domain. These modules can then be compositional combined with a dynamic routing strategy, enabling highly flexible domain adaptation in the retrieval part. Extensive evaluation on six zero-shot downstream tasks demonstrates that this approach can surpass DR while utilizing only 2\% of the training parameters, paving the way to achieve more flexible dense retrieval in IR. We see it as a promising future direction for applying dense retrieval to various tasks.