Search papers, labs, and topics across Lattice.
This paper introduces Uncertainty-based Iterative Document Sampling (UnIte), a novel approach for unsupervised domain adaptation in information retrieval that enhances the selection of documents for pseudo query generation by incorporating model uncertainty. By filtering documents based on high aleatoric uncertainty and prioritizing those with high epistemic uncertainty, UnIte maximizes the learning utility of the current model. Experimental results on the BEIR dataset demonstrate that UnIte achieves significant improvements in retrieval performance, with gains of +2.45 and +3.49 nDCG@10 using a reduced training sample size of 4k documents.
UnIte reveals that incorporating uncertainty into document sampling can lead to substantial improvements in retrieval performance with fewer training samples.
Unsupervised domain adaptation generalizes neural retrievers to an unseen domain by generating pseudo queries on target domain documents. The quality and efficiency of this adaptation critically depend on which documents are selected for pseudo query generation. The existing document sampling method focuses on diversity but fails to capture model uncertainty. In contrast, we propose **Un**certainty-based **Ite**rative Document Sampling (UnIte) addressing these limitations by (1) filtering documents with high aleatoric uncertainty and (2) prioritizing those with high epistemic uncertainty, maximizing the learning utility of the current model. We conducted extensive experiments on a large corpus of BEIR with small and large models, showing significant gains of +2.45 and +3.49 nDCG@10 with a smaller training sample size, 4k on average.
