English

Distribution-Aligned Fine-Tuning for Efficient Neural Retrieval

Information Retrieval 2022-11-10 v1

Abstract

Dual-encoder-based neural retrieval models achieve appreciable performance and complement traditional lexical retrievers well due to their semantic matching capabilities, which makes them a common choice for hybrid IR systems. However, these models exhibit a performance bottleneck in the online query encoding step, as the corresponding query encoders are usually large and complex Transformer models. In this paper we investigate heterogeneous dual-encoder models, where the two encoders are separate models that do not share parameters or initializations. We empirically show that heterogeneous dual-encoders are susceptible to collapsing representations, causing them to output constant trivial representations when they are fine-tuned using a standard contrastive loss due to a distribution mismatch. We propose DAFT, a simple two-stage fine-tuning approach that aligns the two encoders in order to prevent them from collapsing. We further demonstrate how DAFT can be used to train efficient heterogeneous dual-encoder models using lightweight query encoders.

Keywords

Cite

@article{arxiv.2211.04942,
  title  = {Distribution-Aligned Fine-Tuning for Efficient Neural Retrieval},
  author = {Jurek Leonhardt and Marcel Jahnke and Avishek Anand},
  journal= {arXiv preprint arXiv:2211.04942},
  year   = {2022}
}
R2 v1 2026-06-28T05:31:17.751Z