Multivector retrieval models achieve state-of-the-art effectiveness through fine-grained token-level representations, but their deployment incurs substantial computational and memory costs. Current solutions---based on the well-known κ-means clustering algorithm---group similar vectors together to enable both effective compression and efficient retrieval. However, standard κ-means scales poorly with the number of clusters and dataset size, and favours frequent tokens during training while underrepresenting rare, discriminative ones. In this work, we introduce Tachiom, a multivector retrieval system that exploits token-level structure to significantly accelerate both clustering and retrieval. By accounting for tokens' distribution during centroid allocation, Tachiom easily scales to millions of centroids, enabling highly accurate document scoring using only centroids, avoiding expensive token-level computation. Tachiom combines a graph-based index over centroids with an optimized Product Quantization layout for efficient final scoring. Experiments on Ms Marco-v1 and LoTTE show that Tachiom achieves up to 247× faster clustering than κ-means and up to 9.8× retrieval speedup over state-of-the-art systems while maintaining comparable or superior effectiveness.

State-of-the-art Learned Sparse Retrieval (LSR) models, such as \splade, typically employ a Language Modeling (LM) head to project latent hidden states into a lexically-anchored logit matrix. This intermediate matrix is subsequently transformed into a sparse lexical representation through element-wise operations (ReLU, łogp) and max-pooling over the sequence dimension. Despite its effectiveness, the LM head creates a massive memory bottleneck due to the sheer size of the vocabulary (V ), which can range from 30,000 to over 250,000 tokens in recent models. Materializing this matrix creates a significant memory bottleneck, limiting model scaling. The resulting I/O overhead between operators further throttles throughput and runtime performance. In this paper, we propose SPARTON, a fast---memory-efficient---Triton kernel tailored for the LM head in LSR models. SPARTON utilizes a fused approach that integrates the tiled matrix multiplication, ReLU, Log1P, and max-reduction into a single GPU kernel. By performing an early online reduction directly on raw logit tiles, SPARTON avoids materializing the full logit matrix in memory. Our experiments demonstrate that the SPARTON kernel, in isolation, achieves up to a 4.8× speedup and an order-of-magnitude reduction in peak memory usage compared to PyTorch baselines. Integrated into SPLADE (|V | ≈ 30k ), SPARTON enables a 33% larger batch size and 14% faster training with no effectiveness loss. On a multilingual backbone (|V | ≈ 250k ), these gains jump to a 26× larger batch size and 2.5× faster training.

Approximate Nearest Neighbors (ANN) search is a core task in Information Retrieval. However, the high computational demands and reliance on expensive infrastructures limit broader contributions to ANN research. Enabling efficient and effective ANN search on low-resource devices would allow researchers in low-income countries to participate in the ANN community, thereby democratizing the field. Despite its potential, the IR literature offers little work on the feasibility of ANN search under resource constraints. In this proposal, we explore efficient solutions for large-scale ANN search on low-resource devices. We report a preliminary experimentation highlighting current limitations and outlining future challenges.

Pre-trained language models have been widely exploited to learn dense representations of documents and queries for information retrieval. While previous efforts have primarily focused on improving effectiveness and user satisfaction, response time remains a critical bottleneck of conversational search systems. To address this, we exploit the topical locality inherent in conversational queries, i.e., the tendency of queries within a conversation to focus on related topics. By leveraging query embedding similarities, we dynamically restrict the search space to semantically relevant document clusters, reducing computational complexity without compromising retrieval quality. We evaluate our approach on the TREC CAsT, 2019 and 2020 datasets using multiple embedding models and vector indexes, achieving improvements in processing speed of up to 10.3X with little loss in performance (4.3X without any loss). Our results show that the proposed system effectively handles complex, multi-turn queries with high precision and efficiency, offering a practical solution for real-time conversational search.

Learned Sparse Retrieval (LSR) is an effective IR approach that exploits pre-trained language models for encoding text into a learned bag of words. Several efforts in the literature have shown that sparsity is key to enabling a good trade-off between the efficiency and effectiveness of the query processor. To induce the right degree of sparsity, researchers typically use regularization techniques when training LSR models. Recently, new efficient-inverted index-based-retrieval engines have been proposed, leading to a natural question: has the role of regularization changed in training LSR models? In this paper, we conduct an extended evaluation of regularization approaches for LSR where we discuss their effectiveness, efficiency, and out-of-domain generalization capabilities. We first show that regularization can be relaxed to produce more effective LSR en- coders. We also show that query encoding is now the bottleneck limiting the overall query processor performance. To remove this bottleneck, we advance the state-of-the-art of inference-free LSR by proposing Learned Inference-free Retrieval (Li-Lsr). At training time, Li-Lsr learns a score for each token, casting the query encoding step into a seamless table lookup. Our approach yields state-of-the-art effectiveness for both in-domain and out-of-domain evaluation,surpassing Splade-v3-Doc by 1 point of mRR@10 on MsMarco and 1.8 points of nDCG@10 on Beir.

Learned Sparse Retrieval (LSR) techniques use neural machinery to represent queries and documents as learned bags of words. In contrast with other neural retrieval techniques, such as generative retrieval and dense retrieval, LSR has been shown to be a remarkably robust, transferable, and efficient family of methods for retrieving high-quality search results. This half-day tutorial aims to provide an extensive overview of LSR, ranging from its fundamentals to the latest emerging techniques. By the end of the tutorial, attendees will be familiar with the important design decisions of an LSR system, know how to apply them to text and other modalities, and understand the latest techniques for retrieving with them efficiently. Website: https://lsr-tutorial.github.io

Quantization and pruning are two effective Deep Neural Network model compression methods. In this paper, we propose Automatic Prune Binarization ( APB ), a novel compression technique combining quantization with pruning. APB enhances the representational capability of binary networks using a few full-precision weights. Our technique jointly maximizes the accuracy of the network while minimizing its memory impact by deciding whether each weight should be binarized or kept in full precision. We show how to efficiently perform a forward pass through layers compressed using APB by decomposing it into a binary and a sparse-dense matrix multiplication. Moreover, we design two novel efficient algorithms for extremely quantized matrix multiplication on CPU, leveraging highly efficient bitwise operations. The proposed algorithms are 6.9× and 1.5× faster than available state-of-the-art solutions. We extensively evaluate APB on two widely adopted model compression datasets, namely CIFAR-10 and ImageNet. APB shows to deliver better accuracy/memory trade-off compared to state-of-the-art methods based on i) quantization, ii) pruning, and iii) a combination of pruning and quantization. APB also outperforms quantization in the accuracy/efficiency trade-off, being up to 2× faster than the 2-bits quantized model with no loss in accuracy. • A network compression method ( APB ) that combines pruning and quantization. • Two matrix multiplication algorithms for extremely low-bit operands. • APB reduces the network size by one order of magnitude with no accuracy loss.

Learned sparse representations form an effective and interpretable class of embeddings for text retrieval. While exact top-k retrieval over such embeddings faces efficiency challenges, a recent algorithm called Seismic has enabled remarkably fast, highly-accurate approximate retrieval. Seismic statically prunes inverted lists, organizes each list into geometrically-cohesive blocks, and augments each block with a summary vector. At query time, each inverted list associated with a query term is traversed one block at a time in an arbitrary order, with the inner product between the query and summaries determining if a block must be evaluated. When a block is deemed promising, its documents are fully evaluated with a forward index. Seismic is one to two orders of magnitude faster than state-of-the-art inverted index-based solutions and significantly outperforms the winning graph-based submissions to the BigANN 2023 Challenge. In this work, we speed up Seismic further by introducing two innovations to its query processing subroutine. First, we traverse blocks in order of importance, rather than arbitrarily. Second, we take the list of documents retrieved by Seismic and expand it to include the neighbors of each document using an offline k-regular nearest neighbor graph; the expanded list is then ranked to produce the final top-k set. Experiments on two public datasets show that our extension, named SeismicWave, can reach almost-exact accuracy levels and is up to 2.2x faster than Seismic.

Recent studies in Learning to Rank (LtR) have shown the possibility of effectively distilling a neural network from an ensemble of regression trees. This fully enables the use of neural-based ranking models in query processors of modern Web search engines. Nevertheless, ensembles of regression trees outperform neural models both in terms of efficiency and effectiveness on CPU. In this paper, we propose a framework to design and train neural networks outperforming ensembles of regression trees. After distilling the networks from tree-based models, we exploit an efficiency-oriented pruning technique that works by sparsifying the most computationally intensive layers of the model. Moreover, we develop inference time predictors, which help devise neural network architectures that match the desired efficiency requirements. Comprehensive experiments on two public learning-to-rank datasets show that the neural networks produced with our novel approach are competitive in terms of effectiveness-efficiency trade-off when compared with tree-based ensembles by providing up to 4x inference time speed-up without degradation of the ranking quality.

Learned sparse representations form an attractive class of contextual embeddings for text retrieval. That is so because they are effective models of relevance and are interpretable by design. Despite their apparent compatibility with inverted indexes, however, retrieval over sparse embeddings remains challenging. That is due to the distributional differences between learned embeddings and term frequency-based lexical models of relevance such as BM25. Recognizing this challenge, a great deal of research has gone into, among other things, designing retrieval algorithms tailored to the properties of learned sparse representations, including approximate retrieval systems. In fact, this task featured prominently in the latest BigANN Challenge at NeurIPS 2023, where approximate algorithms were evaluated on a large benchmark dataset by throughput and recall. In this work, we propose a novel organization of the inverted index that enables fast yet effective approximate retrieval over learned sparse embeddings. Our approach organizes inverted lists into geometrically-cohesive blocks, each equipped with a summary vector. During query processing, we quickly determine if a block must be evaluated using the summaries. As we show experimentally, single-threaded query processing using our method, Seismic, reaches sub-millisecond per-query latency on various sparse embeddings of the MS MARCO dataset while maintaining high recall. Our results indicate that Seismic is one to two orders of magnitude faster than state-of-the-art inverted index-based solutions and further outperforms the winning (graph-based) submissions to the BigANN Challenge by a significant margin.

Recent studies in Learning to Rank have shown the possibility to effectively distill a neural network from an ensemble of regression trees. This result leads neural networks to become a natural competitor of tree-based ensembles on the ranking task. Nevertheless, ensembles of regression trees outperform neural models both in terms of efficiency and effectiveness, particularly when scoring on CPU. In this paper, we propose an approach for speeding up neural scoring time by applying a combination of Distillation, Pruning and Fast Matrix multiplication. We employ knowledge distillation to learn shallow neural networks from an ensemble of regression trees. Then, we exploit an efficiency-oriented pruning technique that performs a sparsification of the most computationally-intensive layers of the neural network that is then scored with optimized sparse matrix multiplication. Moreover, by studying both dense and sparse high performance matrix multiplication, we develop a scoring time prediction model which helps in devising neural network architectures that match the desired efficiency requirements. Comprehensive experiments on two public learning-to-rank datasets show that neural networks produced with our novel approach are competitive at any point of the effectiveness-efficiency trade-off when compared with tree-based ensembles, providing up to 4x scoring time speed-up without affecting the ranking quality.