From Distillation to Hard Negative Sampling: Making Sparse Neural IR Models More Effective
This work addresses the training gap for sparse retrievers, offering incremental improvements for IR systems by adapting methods from dense models.
The paper tackled the problem of improving sparse neural information retrieval models by applying training enhancements like distillation and hard-negative mining, achieving state-of-the-art results in in-domain and zero-shot settings.
Neural retrievers based on dense representations combined with Approximate Nearest Neighbors search have recently received a lot of attention, owing their success to distillation and/or better sampling of examples for training -- while still relying on the same backbone architecture. In the meantime, sparse representation learning fueled by traditional inverted indexing techniques has seen a growing interest, inheriting from desirable IR priors such as explicit lexical matching. While some architectural variants have been proposed, a lesser effort has been put in the training of such models. In this work, we build on SPLADE -- a sparse expansion-based retriever -- and show to which extent it is able to benefit from the same training improvements as dense models, by studying the effect of distillation, hard-negative mining as well as the Pre-trained Language Model initialization. We furthermore study the link between effectiveness and efficiency, on in-domain and zero-shot settings, leading to state-of-the-art results in both scenarios for sufficiently expressive models.