Recent work on neural sparse retrieval has demonstrated strong relevance by leveraging Large Language Models (LLMs) for semantic term expansion. However, learned models paired with previous sparsification techniques still yield overly long document and query vectors partly due to a large LLM vocabulary, imposing a serious challenge to retrieval time and space efficiency. This paper proposes a scheme for optimizing model sparsity through a synergy of adaptive strategies, including learnable soft top-??, per-term thresholding, and FLOPs regularization to increase the sparsity of query and document vectors. Experimental results with Lion-SP model on the MS MARCO and BEIR datasets demonstrate that the proposed scheme can outperform the baselines by significantly reducing the average query and document lengths. Our scheme can achieve much shorter retrieval latency and lower storage cost while maintaining highly competitive relevance.

This paper proposes superblock pruning (SP) during top-k online document retrieval for learned sparse representations. SP structures the sparse index as a set of superblocks on a sequence of document blocks and conducts a superblock-level selection to decide if some superblocks can be pruned before visiting their child blocks. SP generalizes the previous flat block or cluster-based pruning, allowing the early detection of groups of documents that cannot or are less likely to appear in the final top-k list. SP can accelerate sparse retrieval in a rank-safe or approximate manner under a high-relevance competitiveness constraint. Our experiments show that the proposed scheme significantly outperforms state-of-the-art baselines on MS MARCO passages on a single-threaded CPU.

Low-cost retrieval is crucial for document search on resource-limited computing platforms. This paper presents a staged sparse-to-dense retrieval framework that substitutes expensive dense query encoding with a dense pseudo-query (DPQ), an approximation derived solely from sparse retrieval results. DPQ scheme employs a simple, rank-aware weighting to combine corresponding dense representations of top sparse results, providing an opportunity to efficiently leverage an expensive but expressive LLM or BERT-based dense model without requiring GPUs. The evaluation demonstrates that DPQ-based retrieval runs fast on an affordable platform and outperforms several low-cost baselines in zero-shot retrieval.

This paper revisits dynamic pruning through rank score thresholding in cluster-based sparse retrieval to skip the index partially at cluster and document levels during inference.It proposes a two-parameter pruning control scheme called ASC with a probabilistic guarantee on rank-safeness competitiveness.ASC uses cluster-level maximum weight segmentation to improve accuracy of rank score bound estimation and threshold-driven pruning, and is targeted for speeding up retrieval applications requiring high relevance competitiveness.The experiments with MS MARCO and BEIR show that ASC improves the accuracy and safeness of pruning for better relevance while delivering a low latency on a single-threaded CPU.

Previous work has demonstrated the potential to combine document rankings from dense and sparse retrievers for higher relevance effectiveness. This paper proposes a cluster-based partial dense retrieval scheme guided by sparse retrieval results to optimize fusion between dense and sparse retrieval at a low space and CPU-time cost while retaining a competitive relevance. This scheme exploits the overlap of sparse retrieval results and document embedding clusters, and judiciously selects a limited number of clusters to probabilistically guarantee the inclusion of top sparse results. This paper provides an evaluation of this scheme on its in-domain and zero-shot retrieval performance for the MS MARCO and BEIR datasets.

Transformer-based retrieval and reranking models for text document search are often refined through knowledge distillation together with contrastive learning. A tight distribution matching between the teacher and student models can be hard as over-calibration may degrade training effectiveness when a teacher does not perform well. This paper contrastively reweights KL divergence terms to prioritize the alignment between a student and a teacher model for proper separation of positive and negative documents. This paper analyzes and evaluates the proposed loss function on the MS MARCO and BEIR datasets to demonstrate its effectiveness in improving the relevance of tested student models.

Learned sparse document representations using a transformer-based neural model has been found to be attractive in both relevance effectiveness and time efficiency. This paper describes a representation sparsification scheme based on hard and soft thresholding with an inverted index approximation for faster SPLADE-based document retrieval. It provides analytical and experimental results on the impact of this learnable hybrid thresholding scheme.