An Efficiency Study for SPLADE Models

TL;DR

This paper investigates methods to enhance SPLADE model efficiency, reducing latency and computational costs while maintaining competitive retrieval performance, and introduces techniques to match traditional BM25 latency under similar constraints.

Contribution

The paper proposes novel techniques including L1 regularization, encoder separation, FLOPS-regularized middle-training, and faster encoders to significantly improve SPLADE efficiency and performance.

Findings

Achieved similar latency to BM25 under the same computing constraints.

Increased in-domain retrieval performance metrics.

First neural models to match traditional retrieval latency with minimal performance loss.

Abstract

Latency and efficiency issues are often overlooked when evaluating IR models based on Pretrained Language Models (PLMs) in reason of multiple hardware and software testing scenarios. Nevertheless, efficiency is an important part of such systems and should not be overlooked. In this paper, we focus on improving the efficiency of the SPLADE model since it has achieved state-of-the-art zero-shot performance and competitive results on TREC collections. SPLADE efficiency can be controlled via a regularization factor, but solely controlling this regularization has been shown to not be efficient enough. In order to reduce the latency gap between SPLADE and traditional retrieval systems, we propose several techniques including L1 regularization for queries, a separation of document/query encoders, a FLOPS-regularized middle-training, and the use of faster query encoders. Our benchmark demonstrates that we can drastically improve the efficiency of these models while increasing the performance metrics on in-domain data. To our knowledge, {we propose the first neural models that, under the same computing constraints, \textit{achieve similar latency (less than 4ms difference) as traditional BM25}, while having \textit{similar performance (less than 10\% MRR@10 reduction)} as the state-of-the-art single-stage neural rankers on in-domain data}.