Zero-Shot Dense Retrieval with Momentum Adversarial Domain Invariant Representations

TL;DR

This paper introduces MoDIR, a momentum adversarial training method that enhances zero-shot domain generalization in dense retrieval models by learning domain-invariant representations, significantly improving performance across multiple datasets.

Contribution

The paper proposes a novel momentum adversarial domain invariant learning approach (MoDIR) to improve zero-shot generalization of dense retrieval models across diverse domains.

Findings

MoDIR outperforms baselines on 10+ BEIR datasets in zero-shot settings.

Achieves over 10% relative gains on sensitive datasets.

Demonstrates robustness in domain transfer without target domain labels.

Abstract

Dense retrieval (DR) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search. This requires strong locality properties from the representation space, i.e, the close allocations of each small group of relevant texts, which are hard to generalize to domains without sufficient training data. In this paper, we aim to improve the generalization ability of DR models from source training domains with rich supervision signals to target domains without any relevant labels, in the zero-shot setting. To achieve that, we propose Momentum adversarial Domain Invariant Representation learning (MoDIR), which introduces a momentum method in the DR training process to train a domain classifier distinguishing source versus target, and then adversarially updates the DR encoder to learn domain invariant representations. Our experiments show that MoDIR robustly outperforms its baselines on 10+ ranking datasets from the BEIR benchmark in the zero-shot setup, with more than 10% relative gains on datasets with enough sensitivity for DR models' evaluation. Source code of this paper will be released.