Showing 1–50 of 82 results for author: Nardini, M

Efficient Multivector Retrieval with Token-Aware Clustering and Hierarchical Indexing

Authors: Silvio Martinico, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini

Abstract: 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 k-means clustering algorithm, group similar vectors together to enable both effective compression and efficient retrieval. However, standard k-means scales poorly with… ▽ More 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 k-means clustering algorithm, group similar vectors together to enable both effective compression and efficient retrieval. However, standard k-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-MARCOv1 and LoTTE show that TACHIOM achieves up to $247\times$ faster clustering than k-means and up to $9.8\times$ retrieval speedup over state-of-the-art systems while maintaining comparable or superior effectiveness. △ Less

Submitted 30 April, 2026; originally announced April 2026.

Comments: 6 pages, 2 figures, SIGIR 2026

Sparton: Fast and Memory-Efficient Triton Kernel for Learned Sparse Retrieval

Authors: Thong Nguyen, Cosimo Rulli, Franco Maria Nardini, Rossano Venturini, Andrew Yates

Abstract: 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, Log1P) and max-pooling over the sequence dimension. Despite its effectiveness, t… ▽ More 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, Log1P) 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 4.8x 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 26x larger batch size and 2.5x faster training. △ Less

Submitted 26 March, 2026; originally announced March 2026.

Forward Index Compression for Learned Sparse Retrieval

Authors: Sebastian Bruch, Martino Fontana, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini

Abstract: Text retrieval using learned sparse representations of queries and documents has, over the years, evolved into a highly effective approach to search. It is thanks to recent advances in approximate nearest neighbor search-with the emergence of highly efficient algorithms such as the inverted index-based Seismic and the graph-based Hnsw-that retrieval with sparse representations became viable in pra… ▽ More Text retrieval using learned sparse representations of queries and documents has, over the years, evolved into a highly effective approach to search. It is thanks to recent advances in approximate nearest neighbor search-with the emergence of highly efficient algorithms such as the inverted index-based Seismic and the graph-based Hnsw-that retrieval with sparse representations became viable in practice. In this work, we scrutinize the efficiency of sparse retrieval algorithms and focus particularly on the size of a data structure that is common to all algorithmic flavors and that constitutes a substantial fraction of the overall index size: the forward index. In particular, we seek compression techniques to reduce the storage footprint of the forward index without compromising search quality or inner product computation latency. In our examination with various integer compression techniques, we report that StreamVByte achieves the best trade-off between memory footprint, retrieval accuracy, and latency. We then improve StreamVByte by introducing DotVByte, a new algorithm tailored to inner product computation. Experiments on MsMarco show that our improvements lead to significant space savings while maintaining retrieval efficiency. △ Less

Submitted 5 February, 2026; originally announced February 2026.

Multivector Reranking in the Era of Strong First-Stage Retrievers

Authors: Silvio Martinico, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini

Abstract: Learned multivector representations power modern search systems with strong retrieval effectiveness, but their real-world use is limited by the high cost of exhaustive token-level retrieval. Therefore, most systems adopt a \emph{gather-and-refine} strategy, where a lightweight gather phase selects candidates for full scoring. However, this approach requires expensive searches over large token-leve… ▽ More Learned multivector representations power modern search systems with strong retrieval effectiveness, but their real-world use is limited by the high cost of exhaustive token-level retrieval. Therefore, most systems adopt a \emph{gather-and-refine} strategy, where a lightweight gather phase selects candidates for full scoring. However, this approach requires expensive searches over large token-level indexes and often misses the documents that would rank highest under full similarity. In this paper, we reproduce several state-of-the-art multivector retrieval methods on two publicly available datasets, providing a clear picture of the current multivector retrieval field and observing the inefficiency of token-level gathering. Building on top of that, we show that replacing the token-level gather phase with a single-vector document retriever -- specifically, a learned sparse retriever (LSR) -- produces a smaller and more semantically coherent candidate set. This recasts the gather-and-refine pipeline into the well-established two-stage retrieval architecture. As retrieval latency decreases, query encoding with two neural encoders becomes the dominant computational bottleneck. To mitigate this, we integrate recent inference-free LSR methods, demonstrating that they preserve the retrieval effectiveness of the dual-encoder pipeline while substantially reducing query encoding time. Finally, we investigate multiple reranking configurations that balance efficiency, memory, and effectiveness, and we introduce two optimization techniques that prune low-quality candidates early. Empirical results show that these techniques improve retrieval efficiency by up to 1.8$\times$ with no loss in quality. Overall, our two-stage approach achieves over $24\times$ speedup over the state-of-the-art multivector retrieval systems, while maintaining comparable or superior retrieval quality. △ Less

Submitted 16 January, 2026; v1 submitted 8 January, 2026; originally announced January 2026.

Comments: 17 pages, 2 figures, ECIR 2026

Exact Nearest-Neighbor Search on Energy-Efficient FPGA Devices

Authors: Patrizio Dazzi, William Guglielmo, Franco Maria Nardini, Raffaele Perego, Salvatore Trani

Abstract: This paper investigates the usage of FPGA devices for energy-efficient exact kNN search in high-dimension latent spaces. This work intercepts a relevant trend that tries to support the increasing popularity of learned representations based on neural encoder models by making their large-scale adoption greener and more inclusive. The paper proposes two different energy-efficient solutions adopting t… ▽ More This paper investigates the usage of FPGA devices for energy-efficient exact kNN search in high-dimension latent spaces. This work intercepts a relevant trend that tries to support the increasing popularity of learned representations based on neural encoder models by making their large-scale adoption greener and more inclusive. The paper proposes two different energy-efficient solutions adopting the same FPGA low-level configuration. The first solution maximizes system throughput by processing the queries of a batch in parallel over a streamed dataset not fitting into the FPGA memory. The second minimizes latency by processing each kNN incoming query in parallel over an in-memory dataset. Reproducible experiments on publicly available image and text datasets show that our solution outperforms state-of-the-art CPU-based competitors regarding throughput, latency, and energy consumption. Specifically, experiments show that the proposed FPGA solutions achieve the best throughput in terms of queries per second and the best-observed latency with scale-up factors of up to 16.6X. Similar considerations can be made regarding energy efficiency, where results show that our solutions can achieve up to 11.9X energy saving w.r.t. strong CPU-based competitors. △ Less

Submitted 19 October, 2025; originally announced October 2025.

Blending Learning to Rank and Dense Representations for Efficient and Effective Cascades

Authors: Franco Maria Nardini, Raffaele Perego, Nicola Tonellotto, Salvatore Trani

Abstract: We investigate the exploitation of both lexical and neural relevance signals for ad-hoc passage retrieval. Our exploration involves a large-scale training dataset in which dense neural representations of MS-MARCO queries and passages are complemented and integrated with 253 hand-crafted lexical features extracted from the same corpus. Blending of the relevance signals from the two different groups… ▽ More We investigate the exploitation of both lexical and neural relevance signals for ad-hoc passage retrieval. Our exploration involves a large-scale training dataset in which dense neural representations of MS-MARCO queries and passages are complemented and integrated with 253 hand-crafted lexical features extracted from the same corpus. Blending of the relevance signals from the two different groups of features is learned by a classical Learning-to-Rank (LTR) model based on a forest of decision trees. To evaluate our solution, we employ a pipelined architecture where a dense neural retriever serves as the first stage and performs a nearest-neighbor search over the neural representations of the documents. Our LTR model acts instead as the second stage that re-ranks the set of candidates retrieved by the first stage to enhance effectiveness. The results of reproducible experiments conducted with state-of-the-art dense retrievers on publicly available resources show that the proposed solution significantly enhances the end-to-end ranking performance while relatively minimally impacting efficiency. Specifically, we achieve a boost in nDCG@10 of up to 11% with an increase in average query latency of only 4.3%. This confirms the advantage of seamlessly combining two distinct families of signals that mutually contribute to retrieval effectiveness. △ Less

Submitted 18 October, 2025; originally announced October 2025.

Efficient Sketching and Nearest Neighbor Search Algorithms for Sparse Vector Sets

Authors: Sebastian Bruch, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini

Abstract: Sparse embeddings of data form an attractive class due to their inherent interpretability: Every dimension is tied to a term in some vocabulary, making it easy to visually decipher the latent space. Sparsity, however, poses unique challenges for Approximate Nearest Neighbor Search (ANNS) which finds, from a collection of vectors, the k vectors closest to a query. To encourage research on this unde… ▽ More Sparse embeddings of data form an attractive class due to their inherent interpretability: Every dimension is tied to a term in some vocabulary, making it easy to visually decipher the latent space. Sparsity, however, poses unique challenges for Approximate Nearest Neighbor Search (ANNS) which finds, from a collection of vectors, the k vectors closest to a query. To encourage research on this underexplored topic, sparse ANNS featured prominently in a BigANN Challenge at NeurIPS 2023, where approximate algorithms were evaluated on large benchmark datasets by throughput and accuracy. In this work, we introduce a set of novel data structures and algorithmic methods, a combination of which leads to an elegant, effective, and highly efficient solution to sparse ANNS. Our contributions range from a theoretically-grounded sketching algorithm for sparse vectors to reduce their effective dimensionality while preserving inner product-induced ranks; a geometric organization of the inverted index; and the blending of local and global information to improve the efficiency and efficacy of ANNS. Empirically, our final algorithm, dubbed Seismic, reaches sub-millisecond per-query latency with high accuracy on a large-scale benchmark dataset using a single CPU. △ Less

Submitted 29 September, 2025; originally announced September 2025.

Early-Exit Graph Neural Networks

Authors: Andrea Giuseppe Di Francesco, Maria Sofia Bucarelli, Franco Maria Nardini, Raffaele Perego, Nicola Tonellotto, Fabrizio Silvestri

Abstract: Early-exit mechanisms allow deep neural networks to stop inference once prediction confidence is high, reducing latency and energy on easy inputs while retaining full-depth accuracy on harder ones. Similarly, adding early exit mechanisms to Graph Neural Networks (GNNs), the go-to models for graph-structured data, allows for dynamic trading depth for confidence on simple graphs while maintaining fu… ▽ More Early-exit mechanisms allow deep neural networks to stop inference once prediction confidence is high, reducing latency and energy on easy inputs while retaining full-depth accuracy on harder ones. Similarly, adding early exit mechanisms to Graph Neural Networks (GNNs), the go-to models for graph-structured data, allows for dynamic trading depth for confidence on simple graphs while maintaining full-depth accuracy on harder ones to capture intricate relationships. Yet, their potential in deep GNNs, where over-smoothing, over-squashing or more generally vanishing gradients prevent these model to properly learn, remains largely unexplored. To address this, we introduce Symmetric-Anti-Symmetric GNNs (SAS-GNN), whose symmetry-based inductive biases yield stable intermediate representations that support safe early exits. Building on this backbone, we propose Early-Exit GNNs (EEGNNs), which attach confidence-aware exit neural heads which are trainable end-to-end based on the task objective, enabling on-the-fly termination at node or graph level. Experiments show that EEGNNs learn task-driven exit strategies, while achieving competitive results on heterophilic graphs and long-range tasks. Even when not outperforming the strongest baselines, EEGNNs consistently deliver favorable accuracy-efficiency trade-offs thanks to their adaptive and parameter-efficient design. We plan to release the code to reproduce our experiments. △ Less

Submitted 3 February, 2026; v1 submitted 23 May, 2025; originally announced May 2025.

Comments: 49 pages, 26 figures. Under review

Effective Inference-Free Retrieval for Learned Sparse Representations

Authors: Franco Maria Nardini, Thong Nguyen, Cosimo Rulli, Rossano Venturini, Andrew Yates

Abstract: 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… ▽ More 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 encoders. 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 MS MARCO and 1.8 points of nDCG@10 on BEIR. △ Less

Submitted 30 April, 2025; originally announced May 2025.

Efficient Conversational Search via Topical Locality in Dense Retrieval

Authors: Cristina Ioana Muntean, Franco Maria Nardini, Raffaele Perego, Guido Rocchietti, Cosimo Rulli

Abstract: 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… ▽ More 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 10.4X with little loss in performance (4.4X without any loss). Our results show that the proposed system effectively handles complex, multiturn queries with high precision and efficiency, offering a practical solution for real-time conversational search. △ Less

Submitted 30 April, 2025; originally announced April 2025.

Comments: 5 pages, 2 figures, SIGIR 2025

ACM Class: H.3

Investigating the Scalability of Approximate Sparse Retrieval Algorithms to Massive Datasets

Authors: Sebastian Bruch, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini, Leonardo Venuta

Abstract: Learned sparse text embeddings have gained popularity due to their effectiveness in top-k retrieval and inherent interpretability. Their distributional idiosyncrasies, however, have long hindered their use in real-world retrieval systems. That changed with the recent development of approximate algorithms that leverage the distributional properties of sparse embeddings to speed up retrieval. Noneth… ▽ More Learned sparse text embeddings have gained popularity due to their effectiveness in top-k retrieval and inherent interpretability. Their distributional idiosyncrasies, however, have long hindered their use in real-world retrieval systems. That changed with the recent development of approximate algorithms that leverage the distributional properties of sparse embeddings to speed up retrieval. Nonetheless, in much of the existing literature, evaluation has been limited to datasets with only a few million documents such as MSMARCO. It remains unclear how these systems behave on much larger datasets and what challenges lurk in larger scales. To bridge that gap, we investigate the behavior of state-of-the-art retrieval algorithms on massive datasets. We compare and contrast the recently-proposed Seismic and graph-based solutions adapted from dense retrieval. We extensively evaluate Splade embeddings of 138M passages from MsMarco-v2 and report indexing time and other efficiency and effectiveness metrics. △ Less

Submitted 20 January, 2025; originally announced January 2025.

kANNolo: Sweet and Smooth Approximate k-Nearest Neighbors Search

Authors: Leonardo Delfino, Domenico Erriquez, Silvio Martinico, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini

Abstract: Approximate Nearest Neighbors (ANN) search is a crucial task in several applications like recommender systems and information retrieval. Current state-of-the-art ANN libraries, although being performance-oriented, often lack modularity and ease of use. This translates into them not being fully suitable for easy prototyping and testing of research ideas, an important feature to enable. We address t… ▽ More Approximate Nearest Neighbors (ANN) search is a crucial task in several applications like recommender systems and information retrieval. Current state-of-the-art ANN libraries, although being performance-oriented, often lack modularity and ease of use. This translates into them not being fully suitable for easy prototyping and testing of research ideas, an important feature to enable. We address these limitations by introducing kANNolo, a novel research-oriented ANN library written in Rust and explicitly designed to combine usability with performance effectively. kANNolo introduces a fully composable architecture for ANN search that supports both dense and sparse vector representations. It enables researchers to seamlessly mix and match different similarity measures, vector quantization techniques (e.g., Product Quantization), and index structures (e.g., HNSW) within a single unified framework. These functionalities are managed through Rust traits, allowing shared behaviors to be handled abstractly. This abstraction ensures flexibility and facilitates an easy integration of new components. In this work, we detail the architecture of kANNolo and demonstrate that its flexibility does not compromise performance. The experimental analysis shows that kANNolo achieves state-of-the-art performance in terms of speed-accuracy trade-off while allowing fast and easy prototyping, thus making kANNolo a valuable tool for advancing ANN research. Source code available on GitHub: https://github.com/TusKANNy/kannolo. △ Less

Submitted 1 July, 2026; v1 submitted 10 January, 2025; originally announced January 2025.

Comments: 7 pages, 3 figures

Journal ref: Proc. of the 47th European Conference on Information Retrieval (ECIR 2025), Part IV, pp. 400-406

Power- and Fragmentation-aware Online Scheduling for GPU Datacenters

Authors: Francesco Lettich, Emanuele Carlini, Franco Maria Nardini, Raffaele Perego, Salvatore Trani

Abstract: The rise of Artificial Intelligence and Large Language Models is driving increased GPU usage in data centers for complex training and inference tasks, impacting operational costs, energy demands, and the environmental footprint of large-scale computing infrastructures. This work addresses the online scheduling problem in GPU datacenters, which involves scheduling tasks without knowledge of their f… ▽ More The rise of Artificial Intelligence and Large Language Models is driving increased GPU usage in data centers for complex training and inference tasks, impacting operational costs, energy demands, and the environmental footprint of large-scale computing infrastructures. This work addresses the online scheduling problem in GPU datacenters, which involves scheduling tasks without knowledge of their future arrivals. We focus on two objectives: minimizing GPU fragmentation and reducing power consumption. GPU fragmentation occurs when partial GPU allocations hinder the efficient use of remaining resources, especially as the datacenter nears full capacity. A recent scheduling policy, Fragmentation Gradient Descent (FGD), leverages a fragmentation metric to address this issue. Reducing power consumption is also crucial due to the significant power demands of GPUs. To this end, we propose PWR, a novel scheduling policy to minimize power usage by selecting power-efficient GPU and CPU combinations. This involves a simplified model for measuring power consumption integrated into a Kubernetes score plugin. Through an extensive experimental evaluation in a simulated cluster, we show how PWR, when combined with FGD, achieves a balanced trade-off between reducing power consumption and minimizing GPU fragmentation. △ Less

Submitted 23 December, 2024; originally announced December 2024.

Comments: This work has been submitted to the IEEE for possible publication

The GROND gamma-ray burst sample. I. Overview and statistics

Authors: J. Greiner, T. Krühler, J. Bolmer, S. Klose, P. M. J. Afonso, J. Elliott, R. Filgas, J. F. Graham, D. A. Kann, F. Knust, A. Küpcü Yoldaş, M. Nardini, A. M. Nicuesa Guelbenzu, F. Olivares Estay, A. Rossi, P. Schady, T. Schweyer, V. Sudilovsky, K. Varela, P. Wiseman

Abstract: A dedicated gamma-ray burst (GRB) afterglow observing program was performed between 2007 and 2016 with GROND, a seven-channel optical and near-infrared imager at the 2.2m telescope of the Max-Planck Society at ESO/La Silla. In this first of a series of papers, we describe the GRB observing plan, providing first readings of all so far unpublished GRB afterglow measurements and some observing statis… ▽ More A dedicated gamma-ray burst (GRB) afterglow observing program was performed between 2007 and 2016 with GROND, a seven-channel optical and near-infrared imager at the 2.2m telescope of the Max-Planck Society at ESO/La Silla. In this first of a series of papers, we describe the GRB observing plan, providing first readings of all so far unpublished GRB afterglow measurements and some observing statistics. In total, we observed 514 GRBs with GROND, including 434 Swift-detected GRBs, representing 81\% of the observable Swift sample. For GROND-observations within 30 min of the GRB trigger, the optical/NIR afterglow detection rate is 81\% for long- and 57\% for short-duration GRBs. We report the discovery of ten new GRB afterglows plus one candidate, along with redshift estimates (partly improved) for four GRBs and new host detections for seven GRBs. We identify the (already known) afterglow of GRB 140209A as the sixth GRB exhibiting a 2175 Angstroem dust feature. As a side result, we identified two blazars, with one at a redshift of z=3.8 (in the GRB 131209A field). △ Less

Submitted 18 October, 2024; originally announced October 2024.

Comments: 20 pages, 31 figures, accepted for publication in Astronomy & Astrophysics

Rewriting Conversational Utterances with Instructed Large Language Models

Authors: Elnara Galimzhanova, Cristina Ioana Muntean, Franco Maria Nardini, Raffaele Perego, Guido Rocchietti

Abstract: Many recent studies have shown the ability of large language models (LLMs) to achieve state-of-the-art performance on many NLP tasks, such as question answering, text summarization, coding, and translation. In some cases, the results provided by LLMs are on par with those of human experts. These models' most disruptive innovation is their ability to perform tasks via zero-shot or few-shot promptin… ▽ More Many recent studies have shown the ability of large language models (LLMs) to achieve state-of-the-art performance on many NLP tasks, such as question answering, text summarization, coding, and translation. In some cases, the results provided by LLMs are on par with those of human experts. These models' most disruptive innovation is their ability to perform tasks via zero-shot or few-shot prompting. This capability has been successfully exploited to train instructed LLMs, where reinforcement learning with human feedback is used to guide the model to follow the user's requests directly. In this paper, we investigate the ability of instructed LLMs to improve conversational search effectiveness by rewriting user questions in a conversational setting. We study which prompts provide the most informative rewritten utterances that lead to the best retrieval performance. Reproducible experiments are conducted on publicly-available TREC CAST datasets. The results show that rewriting conversational utterances with instructed LLMs achieves significant improvements of up to 25.2% in MRR, 31.7% in Precision@1, 27% in NDCG@3, and 11.5% in Recall@500 over state-of-the-art techniques. △ Less

Submitted 10 October, 2024; originally announced October 2024.

Journal ref: 2023 IEEE/WIC International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT)

Early Exit Strategies for Approximate k-NN Search in Dense Retrieval

Authors: Francesco Busolin, Claudio Lucchese, Franco Maria Nardini, Salvatore Orlando, Raffaele Perego, Salvatore Trani

Abstract: Learned dense representations are a popular family of techniques for encoding queries and documents using high-dimensional embeddings, which enable retrieval by performing approximate k nearest-neighbors search (A-kNN). A popular technique for making A-kNN search efficient is based on a two-level index, where the embeddings of documents are clustered offline and, at query processing, a fixed numbe… ▽ More Learned dense representations are a popular family of techniques for encoding queries and documents using high-dimensional embeddings, which enable retrieval by performing approximate k nearest-neighbors search (A-kNN). A popular technique for making A-kNN search efficient is based on a two-level index, where the embeddings of documents are clustered offline and, at query processing, a fixed number N of clusters closest to the query is visited exhaustively to compute the result set. In this paper, we build upon state-of-the-art for early exit A-kNN and propose an unsupervised method based on the notion of patience, which can reach competitive effectiveness with large efficiency gains. Moreover, we discuss a cascade approach where we first identify queries that find their nearest neighbor within the closest t << N clusters, and then we decide how many more to visit based on our patience approach or other state-of-the-art strategies. Reproducible experiments employing state-of-the-art dense retrieval models and publicly available resources show that our techniques improve the A-kNN efficiency with up to 5x speedups while achieving negligible effectiveness losses. All the code used is available at https://github.com/francescobusolin/faiss_pEE △ Less

Submitted 9 August, 2024; originally announced August 2024.

Comments: 6 pages, published at CIKM 2024

Pairing Clustered Inverted Indexes with kNN Graphs for Fast Approximate Retrieval over Learned Sparse Representations

Authors: Sebastian Bruch, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini

Abstract: 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 ea… ▽ More 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. △ Less

Submitted 8 August, 2024; originally announced August 2024.

Optimistic Query Routing in Clustering-based Approximate Maximum Inner Product Search

Authors: Sebastian Bruch, Aditya Krishnan, Franco Maria Nardini

Abstract: Clustering-based nearest neighbor search is an effective method in which points are partitioned into geometric shards to form an index, with only a few shards searched during query processing to find a set of top-$k$ vectors. Even though the search efficacy is heavily influenced by the algorithm that identifies the shards to probe, it has received little attention in the literature. This work brid… ▽ More Clustering-based nearest neighbor search is an effective method in which points are partitioned into geometric shards to form an index, with only a few shards searched during query processing to find a set of top-$k$ vectors. Even though the search efficacy is heavily influenced by the algorithm that identifies the shards to probe, it has received little attention in the literature. This work bridges that gap by studying routing in clustering-based maximum inner product search. We unpack existing routers and notice the surprising contribution of optimism. We then take a page from the sequential decision making literature and formalize that insight following the principle of ``optimism in the face of uncertainty.'' In particular, we present a framework that incorporates the moments of the distribution of inner products within each shard to estimate the maximum inner product. We then present an instance of our algorithm that uses only the first two moments to reach the same accuracy as state-of-the-art routers such as ScaNN by probing up to $50\%$ fewer points on benchmark datasets. Our algorithm is also space-efficient: we design a sketch of the second moment whose size is independent of the number of points and requires $\mathcal{O}(1)$ vectors per shard. △ Less

Submitted 17 October, 2025; v1 submitted 20 May, 2024; originally announced May 2024.

Efficient Inverted Indexes for Approximate Retrieval over Learned Sparse Representations

Authors: Sebastian Bruch, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini

Abstract: 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 fre… ▽ More 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. △ Less

Submitted 29 April, 2024; originally announced April 2024.

A Learning-to-Rank Formulation of Clustering-Based Approximate Nearest Neighbor Search

Authors: Thomas Vecchiato, Claudio Lucchese, Franco Maria Nardini, Sebastian Bruch

Abstract: A critical piece of the modern information retrieval puzzle is approximate nearest neighbor search. Its objective is to return a set of $k$ data points that are closest to a query point, with its accuracy measured by the proportion of exact nearest neighbors captured in the returned set. One popular approach to this question is clustering: The indexing algorithm partitions data points into non-ove… ▽ More A critical piece of the modern information retrieval puzzle is approximate nearest neighbor search. Its objective is to return a set of $k$ data points that are closest to a query point, with its accuracy measured by the proportion of exact nearest neighbors captured in the returned set. One popular approach to this question is clustering: The indexing algorithm partitions data points into non-overlapping subsets and represents each partition by a point such as its centroid. The query processing algorithm first identifies the nearest clusters -- a process known as routing -- then performs a nearest neighbor search over those clusters only. In this work, we make a simple observation: The routing function solves a ranking problem. Its quality can therefore be assessed with a ranking metric, making the function amenable to learning-to-rank. Interestingly, ground-truth is often freely available: Given a query distribution in a top-$k$ configuration, the ground-truth is the set of clusters that contain the exact top-$k$ vectors. We develop this insight and apply it to Maximum Inner Product Search (MIPS). As we demonstrate empirically on various datasets, learning a simple linear function consistently improves the accuracy of clustering-based MIPS. △ Less

Submitted 17 April, 2024; originally announced April 2024.

Efficient Multi-Vector Dense Retrieval Using Bit Vectors

Authors: Franco Maria Nardini, Cosimo Rulli, Rossano Venturini

Abstract: Dense retrieval techniques employ pre-trained large language models to build a high-dimensional representation of queries and passages. These representations compute the relevance of a passage w.r.t. to a query using efficient similarity measures. In this line, multi-vector representations show improved effectiveness at the expense of a one-order-of-magnitude increase in memory footprint and query… ▽ More Dense retrieval techniques employ pre-trained large language models to build a high-dimensional representation of queries and passages. These representations compute the relevance of a passage w.r.t. to a query using efficient similarity measures. In this line, multi-vector representations show improved effectiveness at the expense of a one-order-of-magnitude increase in memory footprint and query latency by encoding queries and documents on a per-token level. Recently, PLAID has tackled these problems by introducing a centroid-based term representation to reduce the memory impact of multi-vector systems. By exploiting a centroid interaction mechanism, PLAID filters out non-relevant documents, thus reducing the cost of the successive ranking stages. This paper proposes ``Efficient Multi-Vector dense retrieval with Bit vectors'' (EMVB), a novel framework for efficient query processing in multi-vector dense retrieval. First, EMVB employs a highly efficient pre-filtering step of passages using optimized bit vectors. Second, the computation of the centroid interaction happens column-wise, exploiting SIMD instructions, thus reducing its latency. Third, EMVB leverages Product Quantization (PQ) to reduce the memory footprint of storing vector representations while jointly allowing for fast late interaction. Fourth, we introduce a per-document term filtering method that further improves the efficiency of the last step. Experiments on MS MARCO and LoTTE show that EMVB is up to 2.8x faster while reducing the memory footprint by 1.8x with no loss in retrieval accuracy compared to PLAID. △ Less

Submitted 3 April, 2024; originally announced April 2024.

Bridging Dense and Sparse Maximum Inner Product Search

Authors: Sebastian Bruch, Franco Maria Nardini, Amir Ingber, Edo Liberty

Abstract: Maximum inner product search (MIPS) over dense and sparse vectors have progressed independently in a bifurcated literature for decades; the latter is better known as top-$k$ retrieval in Information Retrieval. This duality exists because sparse and dense vectors serve different end goals. That is despite the fact that they are manifestations of the same mathematical problem. In this work, we ask i… ▽ More Maximum inner product search (MIPS) over dense and sparse vectors have progressed independently in a bifurcated literature for decades; the latter is better known as top-$k$ retrieval in Information Retrieval. This duality exists because sparse and dense vectors serve different end goals. That is despite the fact that they are manifestations of the same mathematical problem. In this work, we ask if algorithms for dense vectors could be applied effectively to sparse vectors, particularly those that violate the assumptions underlying top-$k$ retrieval methods. We study IVF-based retrieval where vectors are partitioned into clusters and only a fraction of clusters are searched during retrieval. We conduct a comprehensive analysis of dimensionality reduction for sparse vectors, and examine standard and spherical KMeans for partitioning. Our experiments demonstrate that IVF serves as an efficient solution for sparse MIPS. As byproducts, we identify two research opportunities and demonstrate their potential. First, we cast the IVF paradigm as a dynamic pruning technique and turn that insight into a novel organization of the inverted index for approximate MIPS for general sparse vectors. Second, we offer a unified regime for MIPS over vectors that have dense and sparse subspaces, and show its robustness to query distributions. △ Less

Submitted 16 September, 2023; originally announced September 2023.

Post-hoc Selection of Pareto-Optimal Solutions in Search and Recommendation

Authors: Vincenzo Paparella, Vito Walter Anelli, Franco Maria Nardini, Raffaele Perego, Tommaso Di Noia

Abstract: Information Retrieval (IR) and Recommender Systems (RS) tasks are moving from computing a ranking of final results based on a single metric to multi-objective problems. Solving these problems leads to a set of Pareto-optimal solutions, known as Pareto frontier, in which no objective can be further improved without hurting the others. In principle, all the points on the Pareto frontier are potentia… ▽ More Information Retrieval (IR) and Recommender Systems (RS) tasks are moving from computing a ranking of final results based on a single metric to multi-objective problems. Solving these problems leads to a set of Pareto-optimal solutions, known as Pareto frontier, in which no objective can be further improved without hurting the others. In principle, all the points on the Pareto frontier are potential candidates to represent the best model selected with respect to the combination of two, or more, metrics. To our knowledge, there are no well-recognized strategies to decide which point should be selected on the frontier. In this paper, we propose a novel, post-hoc, theoretically-justified technique, named "Population Distance from Utopia" (PDU), to identify and select the one-best Pareto-optimal solution from the frontier. In detail, PDU analyzes the distribution of the points by investigating how far each point is from its utopia point (the ideal performance for the objectives). The possibility of considering fine-grained utopia points allows PDU to select solutions tailored to individual user preferences, a novel feature we call "calibration". We compare PDU against existing state-of-the-art strategies through extensive experiments on tasks from both IR and RS. Experimental results show that PDU and combined with calibration notably impact the solution selection. Furthermore, the results show that the proposed framework selects a solution in a principled way, irrespective of its position on the frontier, thus overcoming the limits of other strategies. △ Less

Submitted 21 June, 2023; originally announced June 2023.

Neural Network Compression using Binarization and Few Full-Precision Weights

Authors: Franco Maria Nardini, Cosimo Rulli, Salvatore Trani, Rossano Venturini

Abstract: Quantization and pruning are two effective Deep Neural Networks 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… ▽ More Quantization and pruning are two effective Deep Neural Networks 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 6.9x and 1.5x faster than available state-of-the-art solutions. We extensively evaluate APB on two widely adopted model compression datasets, namely CIFAR10 and ImageNet. APB delivers better accuracy/memory trade-off compared to state-of-the-art methods based on i) quantization, ii) pruning, and iii) combination of pruning and quantization. APB outperforms quantization in the accuracy/efficiency trade-off, being up to 2x faster than the 2-bit quantized model with no loss in accuracy. △ Less

Submitted 15 September, 2023; v1 submitted 15 June, 2023; originally announced June 2023.

Comments: 15 pages, 6 figures, 3 tables

ACM Class: I.2.6

Efficient and Effective Tree-based and Neural Learning to Rank

Authors: Sebastian Bruch, Claudio Lucchese, Franco Maria Nardini

Abstract: This monograph takes a step towards promoting the study of efficiency in the era of neural information retrieval by offering a comprehensive survey of the literature on efficiency and effectiveness in ranking, and to a limited extent, retrieval. This monograph was inspired by the parallels that exist between the challenges in neural network-based ranking solutions and their predecessors, decision… ▽ More This monograph takes a step towards promoting the study of efficiency in the era of neural information retrieval by offering a comprehensive survey of the literature on efficiency and effectiveness in ranking, and to a limited extent, retrieval. This monograph was inspired by the parallels that exist between the challenges in neural network-based ranking solutions and their predecessors, decision forest-based learning to rank models, as well as the connections between the solutions the literature to date has to offer. We believe that by understanding the fundamentals underpinning these algorithmic and data structure solutions for containing the contentious relationship between efficiency and effectiveness, one can better identify future directions and more efficiently determine the merits of ideas. We also present what we believe to be important research directions in the forefront of efficiency and effectiveness in retrieval and ranking. △ Less

Submitted 15 May, 2023; originally announced May 2023.

An Approximate Algorithm for Maximum Inner Product Search over Streaming Sparse Vectors

Authors: Sebastian Bruch, Franco Maria Nardini, Amir Ingber, Edo Liberty

Abstract: Maximum Inner Product Search or top-k retrieval on sparse vectors is well-understood in information retrieval, with a number of mature algorithms that solve it exactly. However, all existing algorithms are tailored to text and frequency-based similarity measures. To achieve optimal memory footprint and query latency, they rely on the near stationarity of documents and on laws governing natural lan… ▽ More Maximum Inner Product Search or top-k retrieval on sparse vectors is well-understood in information retrieval, with a number of mature algorithms that solve it exactly. However, all existing algorithms are tailored to text and frequency-based similarity measures. To achieve optimal memory footprint and query latency, they rely on the near stationarity of documents and on laws governing natural languages. We consider, instead, a setup in which collections are streaming -- necessitating dynamic indexing -- and where indexing and retrieval must work with arbitrarily distributed real-valued vectors. As we show, existing algorithms are no longer competitive in this setup, even against naive solutions. We investigate this gap and present a novel approximate solution, called Sinnamon, that can efficiently retrieve the top-k results for sparse real valued vectors drawn from arbitrary distributions. Notably, Sinnamon offers levers to trade-off memory consumption, latency, and accuracy, making the algorithm suitable for constrained applications and systems. We give theoretical results on the error introduced by the approximate nature of the algorithm, and present an empirical evaluation of its performance on two hardware platforms and synthetic and real-valued datasets. We conclude by laying out concrete directions for future research on this general top-k retrieval problem over sparse vectors. △ Less

Submitted 25 January, 2023; originally announced January 2023.

Caching Historical Embeddings in Conversational Search

Authors: Ophir Frieder, Ida Mele, Cristina Ioana Muntean, Franco Maria Nardini, Raffaele Perego, Nicola Tonellotto

Abstract: Rapid response, namely low latency, is fundamental in search applications; it is particularly so in interactive search sessions, such as those encountered in conversational settings. An observation with a potential to reduce latency asserts that conversational queries exhibit a temporal locality in the lists of documents retrieved. Motivated by this observation, we propose and evaluate a client-si… ▽ More Rapid response, namely low latency, is fundamental in search applications; it is particularly so in interactive search sessions, such as those encountered in conversational settings. An observation with a potential to reduce latency asserts that conversational queries exhibit a temporal locality in the lists of documents retrieved. Motivated by this observation, we propose and evaluate a client-side document embedding cache, improving the responsiveness of conversational search systems. By leveraging state-of-the-art dense retrieval models to abstract document and query semantics, we cache the embeddings of documents retrieved for a topic introduced in the conversation, as they are likely relevant to successive queries. Our document embedding cache implements an efficient metric index, answering nearest-neighbor similarity queries by estimating the approximate result sets returned. We demonstrate the efficiency achieved using our cache via reproducible experiments based on TREC CAsT datasets, achieving a hit rate of up to 75% without degrading answer quality. Our achieved high cache hit rates significantly improve the responsiveness of conversational systems while likewise reducing the number of queries managed on the search back-end. △ Less

Submitted 25 November, 2022; originally announced November 2022.

ILMART: Interpretable Ranking with Constrained LambdaMART

Authors: Claudio Lucchese, Franco Maria Nardini, Salvatore Orlando, Raffaele Perego, Alberto Veneri

Abstract: Interpretable Learning to Rank (LtR) is an emerging field within the research area of explainable AI, aiming at developing intelligible and accurate predictive models. While most of the previous research efforts focus on creating post-hoc explanations, in this paper we investigate how to train effective and intrinsically-interpretable ranking models. Developing these models is particularly challen… ▽ More Interpretable Learning to Rank (LtR) is an emerging field within the research area of explainable AI, aiming at developing intelligible and accurate predictive models. While most of the previous research efforts focus on creating post-hoc explanations, in this paper we investigate how to train effective and intrinsically-interpretable ranking models. Developing these models is particularly challenging and it also requires finding a trade-off between ranking quality and model complexity. State-of-the-art rankers, made of either large ensembles of trees or several neural layers, exploit in fact an unlimited number of feature interactions making them black boxes. Previous approaches on intrinsically-interpretable ranking models address this issue by avoiding interactions between features thus paying a significant performance drop with respect to full-complexity models. Conversely, ILMART, our novel and interpretable LtR solution based on LambdaMART, is able to train effective and intelligible models by exploiting a limited and controlled number of pairwise feature interactions. Exhaustive and reproducible experiments conducted on three publicly-available LtR datasets show that ILMART outperforms the current state-of-the-art solution for interpretable ranking of a large margin with a gain of nDCG of up to 8%. △ Less

Submitted 1 June, 2022; originally announced June 2022.

Comments: 5 pages, 3 figures, to be published in SIGIR 2022 proceedings

Distilled Neural Networks for Efficient Learning to Rank

Authors: F. M. Nardini, C. Rulli, S. Trani, R. Venturini

Abstract: 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.… ▽ More 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. △ Less

Submitted 22 February, 2022; originally announced February 2022.

Comments: This work has been submitted to the IEEE for possible publication

An Optimal Algorithm for Finding Champions in Tournament Graphs

Authors: Lorenzo Beretta, Franco Maria Nardini, Roberto Trani, Rossano Venturini

Abstract: A tournament graph is a complete directed graph, which can be used to model a round-robin tournament between $n$ players. In this paper, we address the problem of finding a champion of the tournament, also known as Copeland winner, which is a player that wins the highest number of matches. In detail, we aim to investigate algorithms that find the champion by playing a low number of matches. Solvin… ▽ More A tournament graph is a complete directed graph, which can be used to model a round-robin tournament between $n$ players. In this paper, we address the problem of finding a champion of the tournament, also known as Copeland winner, which is a player that wins the highest number of matches. In detail, we aim to investigate algorithms that find the champion by playing a low number of matches. Solving this problem allows us to speed up several Information Retrieval and Recommender System applications, including question answering, conversational search, etc. Indeed, these applications often search for the champion inducing a round-robin tournament among the players by employing a machine learning model to estimate who wins each pairwise comparison. Our contribution, thus, allows finding the champion by performing a low number of model inferences. We prove that any deterministic or randomized algorithm finding a champion with constant success probability requires $Ω(\ell n)$ comparisons, where $\ell$ is the number of matches lost by the champion. We then present an asymptotically-optimal deterministic algorithm matching this lower bound without knowing $\ell$, and we extend our analysis to three variants of the problem. Lastly, we conduct a comprehensive experimental assessment of the proposed algorithms on a question answering task on public data. Results show that our proposed algorithms speed up the retrieval of the champion up to $13\times$ with respect to the state-of-the-art algorithm that perform the full tournament. △ Less

Submitted 18 April, 2023; v1 submitted 26 November, 2021; originally announced November 2021.

Learning Early Exit Strategies for Additive Ranking Ensembles

Authors: Francesco Busolin, Claudio Lucchese, Franco Maria Nardini, Salvatore Orlando, Raffaele Perego, Salvatore Trani

Abstract: Modern search engine ranking pipelines are commonly based on large machine-learned ensembles of regression trees. We propose LEAR, a novel - learned - technique aimed to reduce the average number of trees traversed by documents to accumulate the scores, thus reducing the overall query response time. LEAR exploits a classifier that predicts whether a document can early exit the ensemble because it… ▽ More Modern search engine ranking pipelines are commonly based on large machine-learned ensembles of regression trees. We propose LEAR, a novel - learned - technique aimed to reduce the average number of trees traversed by documents to accumulate the scores, thus reducing the overall query response time. LEAR exploits a classifier that predicts whether a document can early exit the ensemble because it is unlikely to be ranked among the final top-k results. The early exit decision occurs at a sentinel point, i.e., after having evaluated a limited number of trees, and the partial scores are exploited to filter out non-promising documents. We evaluate LEAR by deploying it in a production-like setting, adopting a state-of-the-art algorithm for ensembles traversal. We provide a comprehensive experimental evaluation on two public datasets. The experiments show that LEAR has a significant impact on the efficiency of the query processing without hindering its ranking quality. In detail, on a first dataset, LEAR is able to achieve a speedup of 3x without any loss in NDCG1@0, while on a second dataset the speedup is larger than 5x with a negligible NDCG@10 loss (< 0.05%). △ Less

Submitted 6 May, 2021; originally announced May 2021.

Comments: 5 pages, 3 figures, ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 21)

ACM Class: H.3.3

Journal ref: 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, Association for Computing Machinery, 2021, 2217-2221

Dynamic Hard Pruning of Neural Networks at the Edge of the Internet

Authors: Lorenzo Valerio, Franco Maria Nardini, Andrea Passarella, Raffaele Perego

Abstract: Neural Networks (NN), although successfully applied to several Artificial Intelligence tasks, are often unnecessarily over-parametrised. In edge/fog computing, this might make their training prohibitive on resource-constrained devices, contrasting with the current trend of decentralising intelligence from remote data centres to local constrained devices. Therefore, we investigate the problem of tr… ▽ More Neural Networks (NN), although successfully applied to several Artificial Intelligence tasks, are often unnecessarily over-parametrised. In edge/fog computing, this might make their training prohibitive on resource-constrained devices, contrasting with the current trend of decentralising intelligence from remote data centres to local constrained devices. Therefore, we investigate the problem of training effective NN models on constrained devices having a fixed, potentially small, memory budget. We target techniques that are both resource-efficient and performance effective while enabling significant network compression. Our Dynamic Hard Pruning (DynHP) technique incrementally prunes the network during training, identifying neurons that marginally contribute to the model accuracy. DynHP enables a tunable size reduction of the final neural network and reduces the NN memory occupancy during training. Freed memory is reused by a \emph{dynamic batch sizing} approach to counterbalance the accuracy degradation caused by the hard pruning strategy, improving its convergence and effectiveness. We assess the performance of DynHP through reproducible experiments on three public datasets, comparing them against reference competitors. Results show that DynHP compresses a NN up to $10$ times without significant performance drops (up to $3.5\%$ additional error w.r.t. the competitors), reducing up to $80\%$ the training memory occupancy. △ Less

Submitted 22 October, 2021; v1 submitted 17 November, 2020; originally announced November 2020.

Query-level Early Exit for Additive Learning-to-Rank Ensembles

Authors: Claudio Lucchese, Franco Maria Nardini, Salvatore Orlando, Raffaele Perego, Salvatore Trani

Abstract: Search engine ranking pipelines are commonly based on large ensembles of machine-learned decision trees. The tight constraints on query response time recently motivated researchers to investigate algorithms to make faster the traversal of the additive ensemble or to early terminate the evaluation of documents that are unlikely to be ranked among the top-k. In this paper, we investigate the novel p… ▽ More Search engine ranking pipelines are commonly based on large ensembles of machine-learned decision trees. The tight constraints on query response time recently motivated researchers to investigate algorithms to make faster the traversal of the additive ensemble or to early terminate the evaluation of documents that are unlikely to be ranked among the top-k. In this paper, we investigate the novel problem of \textit{query-level early exiting}, aimed at deciding the profitability of early stopping the traversal of the ranking ensemble for all the candidate documents to be scored for a query, by simply returning a ranking based on the additive scores computed by a limited portion of the ensemble. Besides the obvious advantage on query latency and throughput, we address the possible positive impact of query-level early exiting on ranking effectiveness. To this end, we study the actual contribution of incremental portions of the tree ensemble to the ranking of the top-k documents scored for a given query. Our main finding is that queries exhibit different behaviors as scores are accumulated during the traversal of the ensemble and that query-level early stopping can remarkably improve ranking quality. We present a reproducible and comprehensive experimental evaluation, conducted on two public datasets, showing that query-level early exiting achieves an overall gain of up to 7.5% in terms of NDCG@10 with a speedup of the scoring process of up to 2.2x. △ Less

Submitted 30 April, 2020; originally announced April 2020.

Comments: Accepted at SIGIR 2020 (short paper)

MSC Class: 68P20

Training Curricula for Open Domain Answer Re-Ranking

Authors: Sean MacAvaney, Franco Maria Nardini, Raffaele Perego, Nicola Tonellotto, Nazli Goharian, Ophir Frieder

Abstract: In precision-oriented tasks like answer ranking, it is more important to rank many relevant answers highly than to retrieve all relevant answers. It follows that a good ranking strategy would be to learn how to identify the easiest correct answers first (i.e., assign a high ranking score to answers that have characteristics that usually indicate relevance, and a low ranking score to those with cha… ▽ More In precision-oriented tasks like answer ranking, it is more important to rank many relevant answers highly than to retrieve all relevant answers. It follows that a good ranking strategy would be to learn how to identify the easiest correct answers first (i.e., assign a high ranking score to answers that have characteristics that usually indicate relevance, and a low ranking score to those with characteristics that do not), before incorporating more complex logic to handle difficult cases (e.g., semantic matching or reasoning). In this work, we apply this idea to the training of neural answer rankers using curriculum learning. We propose several heuristics to estimate the difficulty of a given training sample. We show that the proposed heuristics can be used to build a training curriculum that down-weights difficult samples early in the training process. As the training process progresses, our approach gradually shifts to weighting all samples equally, regardless of difficulty. We present a comprehensive evaluation of our proposed idea on three answer ranking datasets. Results show that our approach leads to superior performance of two leading neural ranking architectures, namely BERT and ConvKNRM, using both pointwise and pairwise losses. When applied to a BERT-based ranker, our method yields up to a 4% improvement in MRR and a 9% improvement in P@1 (compared to the model trained without a curriculum). This results in models that can achieve comparable performance to more expensive state-of-the-art techniques. △ Less

Submitted 21 May, 2020; v1 submitted 29 April, 2020; originally announced April 2020.

Comments: Accepted at SIGIR 2020 (long)

Efficient Document Re-Ranking for Transformers by Precomputing Term Representations

Authors: Sean MacAvaney, Franco Maria Nardini, Raffaele Perego, Nicola Tonellotto, Nazli Goharian, Ophir Frieder

Abstract: Deep pretrained transformer networks are effective at various ranking tasks, such as question answering and ad-hoc document ranking. However, their computational expenses deem them cost-prohibitive in practice. Our proposed approach, called PreTTR (Precomputing Transformer Term Representations), considerably reduces the query-time latency of deep transformer networks (up to a 42x speedup on web do… ▽ More Deep pretrained transformer networks are effective at various ranking tasks, such as question answering and ad-hoc document ranking. However, their computational expenses deem them cost-prohibitive in practice. Our proposed approach, called PreTTR (Precomputing Transformer Term Representations), considerably reduces the query-time latency of deep transformer networks (up to a 42x speedup on web document ranking) making these networks more practical to use in a real-time ranking scenario. Specifically, we precompute part of the document term representations at indexing time (without a query), and merge them with the query representation at query time to compute the final ranking score. Due to the large size of the token representations, we also propose an effective approach to reduce the storage requirement by training a compression layer to match attention scores. Our compression technique reduces the storage required up to 95% and it can be applied without a substantial degradation in ranking performance. △ Less

Submitted 26 May, 2020; v1 submitted 29 April, 2020; originally announced April 2020.

Comments: Accepted at SIGIR 2020 (long)

Expansion via Prediction of Importance with Contextualization

Authors: Sean MacAvaney, Franco Maria Nardini, Raffaele Perego, Nicola Tonellotto, Nazli Goharian, Ophir Frieder

Abstract: The identification of relevance with little textual context is a primary challenge in passage retrieval. We address this problem with a representation-based ranking approach that: (1) explicitly models the importance of each term using a contextualized language model; (2) performs passage expansion by propagating the importance to similar terms; and (3) grounds the representations in the lexicon,… ▽ More The identification of relevance with little textual context is a primary challenge in passage retrieval. We address this problem with a representation-based ranking approach that: (1) explicitly models the importance of each term using a contextualized language model; (2) performs passage expansion by propagating the importance to similar terms; and (3) grounds the representations in the lexicon, making them interpretable. Passage representations can be pre-computed at index time to reduce query-time latency. We call our approach EPIC (Expansion via Prediction of Importance with Contextualization). We show that EPIC significantly outperforms prior importance-modeling and document expansion approaches. We also observe that the performance is additive with the current leading first-stage retrieval methods, further narrowing the gap between inexpensive and cost-prohibitive passage ranking approaches. Specifically, EPIC achieves a MRR@10 of 0.304 on the MS-MARCO passage ranking dataset with 78ms average query latency on commodity hardware. We also find that the latency is further reduced to 68ms by pruning document representations, with virtually no difference in effectiveness. △ Less

Submitted 20 May, 2020; v1 submitted 29 April, 2020; originally announced April 2020.

Comments: Accepted at SIGIR 2020 (short)

Topic Propagation in Conversational Search

Authors: I. Mele, C. I. Muntean, F. M. Nardini, R. Perego, N. Tonellotto, O. Frieder

Abstract: In a conversational context, a user expresses her multi-faceted information need as a sequence of natural-language questions, i.e., utterances. Starting from a given topic, the conversation evolves through user utterances and system replies. The retrieval of documents relevant to a given utterance in a conversation is challenging due to ambiguity of natural language and to the difficulty of detect… ▽ More In a conversational context, a user expresses her multi-faceted information need as a sequence of natural-language questions, i.e., utterances. Starting from a given topic, the conversation evolves through user utterances and system replies. The retrieval of documents relevant to a given utterance in a conversation is challenging due to ambiguity of natural language and to the difficulty of detecting possible topic shifts and semantic relationships among utterances. We adopt the 2019 TREC Conversational Assistant Track (CAsT) framework to experiment with a modular architecture performing: (i) topic-aware utterance rewriting, (ii) retrieval of candidate passages for the rewritten utterances, and (iii) neural-based re-ranking of candidate passages. We present a comprehensive experimental evaluation of the architecture assessed in terms of traditional IR metrics at small cutoffs. Experimental results show the effectiveness of our techniques that achieve an improvement up to 0.28 (+93%) for P@1 and 0.19 (+89.9%) for nDCG@3 w.r.t. the CAsT baseline. △ Less

Submitted 29 April, 2020; originally announced April 2020.

Comments: 5 pages

The luminous host galaxy, faint supernova and rapid afterglow rebrightening of GRB 100418A

Authors: A. de Ugarte Postigo, C. C. Thoene, K. Bensch, A. J. van der Horst, D. A. Kann, Z. Cano, L. Izzo, P. Goldoni, S. Martin, R. Filgas, P. Schady, J. Gorosabel, I. Bikmaev, M. Bremer, R. Burenin, A. J. Castro-Tirado, S. Covino, J. P. U. Fynbo, D. Garcia-Appadoo, I. de Gregorio-Monsalvo, M. Jelinek, I. Khamitov, A. Kamble, C. Kouveliotou, T. Kruehler , et al. (13 additional authors not shown)

Abstract: Long gamma-ray bursts give us the chance to study both their extreme physics and the star-forming galaxies in which they form. GRB 100418A, at a z = 0.6239, had a bright optical and radio afterglow, and a luminous star-forming host galaxy. This allowed us to study the radiation of the explosion as well as the interstellar medium of the host both in absorption and emission. We collected photometric… ▽ More Long gamma-ray bursts give us the chance to study both their extreme physics and the star-forming galaxies in which they form. GRB 100418A, at a z = 0.6239, had a bright optical and radio afterglow, and a luminous star-forming host galaxy. This allowed us to study the radiation of the explosion as well as the interstellar medium of the host both in absorption and emission. We collected photometric data from radio to X-ray wavelengths to study the evolution of the afterglow and the contribution of a possible supernova and three X-shooter spectra obtained during the first 60 hr. The light curve shows a very fast optical rebrightening, with an amplitude of 3 magnitudes, starting 2.4 hr after the GRB onset. This cannot be explained by a standard external shock model and requires other contributions, such as late central-engine activity. Two weeks after the burst we detect an excess in the light curve consistent with a SN with peak absolute magnitude M_V = -18.5 mag, among the faintest GRB-SNe detected to date. The host galaxy shows two components in emission, with velocities differing by 130 km s^-1, but otherwise having similar properties. While some absorption and emission components coincide, the absorbing gas spans much higher velocities, indicating the presence of gas beyond the star-forming regions. The host has a star-formation rate of 12.2 M_sol yr^-1, a metallicity of 12 + log(O/H) = 8.55 and a mass of 1.6x10^9 M_sol. GRB 100418A is a member of a class of afterglow light curves which show a steep rebrightening in the optical during the first day, which cannot be explained by traditional models. Its very faint associated SN shows that GRB-SNe can have a larger dispersion in luminosities than previously seen. Furthermore, we have obtained a complete view of the host of GRB 100418A owing to its spectrum, which contains a remarkable number of both emission and absorption lines. △ Less

Submitted 23 August, 2018; v1 submitted 11 July, 2018; originally announced July 2018.

Comments: 23 pages, 14 figures, accepted for publication in A&A

Journal ref: A&A 620, A190 (2018)

The environment of the SN-less GRB 111005A at z = 0.0133

Authors: M. Tanga, T. Krühler, P. Schady, S. Klose, J. F. Graham, J. Greiner, D. A. Kann, M. Nardini

Abstract: The collapsar model has proved highly successful in explaining the properties of long gamma-ray bursts (GRBs), with the most direct confirmation being the detection of a supernova (SN) coincident with the majority of nearby long GRBs. Within this model, a long GRB is produced by the core-collapse of a metal-poor, rapidly rotating, massive star. The detection of some long GRBs in metal-rich environ… ▽ More The collapsar model has proved highly successful in explaining the properties of long gamma-ray bursts (GRBs), with the most direct confirmation being the detection of a supernova (SN) coincident with the majority of nearby long GRBs. Within this model, a long GRB is produced by the core-collapse of a metal-poor, rapidly rotating, massive star. The detection of some long GRBs in metal-rich environments, and more fundamentally the three examples of long GRBs (GRB 060505, GRB 060614 and GRB 111005A) with no coincident SN detection down to very deep limits is in strong contention with theoretical expectations. In this paper we present MUSE observations of the host galaxy of GRB 111005A, which is the most recent and compelling example yet of a SN-less, long GRB. At z=0.01326, GRB 111005A is the third closest GRB ever detected, and second closest long duration GRB, enabling the nearby environment to be studied at a resolution of 270 pc. From the analysis of the MUSE data cube, we find GRB 111005A to have occurred within a metal-rich environment with little signs of ongoing star formation. Spectral analysis at the position of the GRB indicates the presence of an old stellar population (tau > 10 Myr), which limits the mass of the GRB progenitor to M_ZAMS<15 Msolar, in direct conflict with the collapsar model. Our deep limits on the presence of any SN emission combined with the environmental conditions at the position of GRB 111005A necessitate the exploration of a novel long GRB formation mechanism that is unrelated to massive stars. △ Less

Submitted 2 May, 2018; v1 submitted 21 August, 2017; originally announced August 2017.

Comments: Now accepted by A&A. Manuscript replaced to match accepted version. Some additional discussion added, and velocity map of the host galaxy now included

Journal ref: A&A 615, A136 (2018)

Sentiment-enhanced Multidimensional Analysis of Online Social Networks: Perception of the Mediterranean Refugees Crisis

Authors: Mauro Coletto, Claudio Lucchese, Cristina Ioana Muntean, Franco Maria Nardini, Andrea Esuli, Chiara Renso, Raffaele Perego

Abstract: We propose an analytical framework able to investigate discussions about polarized topics in online social networks from many different angles. The framework supports the analysis of social networks along several dimensions: time, space and sentiment. We show that the proposed analytical framework and the methodology can be used to mine knowledge about the perception of complex social phenomena. W… ▽ More We propose an analytical framework able to investigate discussions about polarized topics in online social networks from many different angles. The framework supports the analysis of social networks along several dimensions: time, space and sentiment. We show that the proposed analytical framework and the methodology can be used to mine knowledge about the perception of complex social phenomena. We selected the refugee crisis discussions over Twitter as the case study. This difficult and controversial topic is an increasingly important issue for the EU. The raw stream of tweets is enriched with space information (user and mentioned locations), and sentiment (positive vs. negative) w.r.t. refugees. Our study shows differences in positive and negative sentiment in EU countries, in particular in UK, and by matching events, locations and perception it underlines opinion dynamics and common prejudices regarding the refugees. △ Less

Submitted 6 May, 2016; originally announced May 2016.

Multiwavelength analysis of three SNe associated with GRBs observed by GROND

Authors: F. Olivares E., J. Greiner, P. Schady, S. Klose, T. Krühler, A. Rau, S. Savaglio, D. A. Kann, G. Pignata, J. Elliott, A. Rossi, M. Nardini, P. M. J. Afonso, R. Filgas, A. Nicuesa Guelbenzu, S. Schmidl, V. Sudilovsky

Abstract: After the discovery of the first connection between GRBs and SNe almost two decades ago, tens of SN-like rebrightenings have been discovered and about seven solid associations have been spectroscopically confirmed to date. Using GROND optical/NIR data and Swift X-ray/UV data, we estimate the intrinsic extinction, luminosity, and evolution of three SN rebrightenings in GRB afterglow light curves at… ▽ More After the discovery of the first connection between GRBs and SNe almost two decades ago, tens of SN-like rebrightenings have been discovered and about seven solid associations have been spectroscopically confirmed to date. Using GROND optical/NIR data and Swift X-ray/UV data, we estimate the intrinsic extinction, luminosity, and evolution of three SN rebrightenings in GRB afterglow light curves at z~0.5. The SNe 2008hw, 2009nz, and 2010ma exhibit 0.80, 1.15, and 1.78 times the optical (r band) luminosity of SN 1998bw, respectively. While SN 2009nz evolves similarly to SN 1998bw, SNe 2008hw and 2010ma show earlier peak times. The quasi-bolometric light curves were corrected for the contribution of the NIR bands using data available in the literature and blackbody fits. The large luminosity of SN 2010ma (1.4x10^43 erg/s) is confirmed, while SNe 2008hw and 2009nz reached a peak luminosity closer to SN 1998bw. Physical parameters of the SN explosions, such as synthesised nickel mass, ejecta mass, and kinetic energy, are estimated using Arnett's analytic approach, which resulted in nickel masses of around 0.4-0.5 Msun. By means of the a very comprehensive data set, we found that the luminosity and the nickel mass of SNe 2008hw, 2009nz, and 2010ma resembles those of other known GRB-associated SNe. This findings strengthens previous claims of GRB-SNe being brighter than type-Ic SNe unaccompanied by GRBs. △ Less

Submitted 2 February, 2015; originally announced February 2015.

Comments: 11 pages, 9 figures, accepted for publication in Astronomy & Astrophysics, abstract abridged

Journal ref: A&A 577, A44 (2015)

Blazar candidates beyond redshift 4 observed by Swift

Authors: T. Sbarrato, G. Ghisellini, G. Tagliaferri, L. Foschini, M. Nardini, F. Tavecchio, N. Gehrels

Abstract: We have selected SDSS J222032.50+002537.5 and SDSS J142048.01+120545.9 as best blazar candidates out of a complete sample of extremely radio-loud quasars at z>4, with highly massive black holes. We observed them and a third serendipitous candidate with similar features (PMN J2134-0419) in the X-rays with the Swift/XRT telescope, to confirm their blazar nature. We observed strong and hard X-ray flu… ▽ More We have selected SDSS J222032.50+002537.5 and SDSS J142048.01+120545.9 as best blazar candidates out of a complete sample of extremely radio-loud quasars at z>4, with highly massive black holes. We observed them and a third serendipitous candidate with similar features (PMN J2134-0419) in the X-rays with the Swift/XRT telescope, to confirm their blazar nature. We observed strong and hard X-ray fluxes (i.e. $α_X<0.6$, where $F(ν)\proptoν^{-α_X}$ in the 0.3-10keV observed energy range, ~1-40keV rest frame) in all three cases. This allowed us to classify our candidates as real blazars, being characterized by large Lorentz factors (~13) and very small viewing angles (~3deg). All three sources have black hole masses exceeding 10^9Msun and their classification provides intriguing constraints on supermassive black hole formation and evolution models. We confirm our earlier suggestion that there are different formation epochs of extremely massive black holes hosted in jetted (z~4) and non-jetted systems (z~2.5). △ Less

Submitted 16 January, 2015; v1 submitted 1 October, 2014; originally announced October 2014.

Comments: 8 pages, 6 figures, published on MNRAS

GRB 120422A/SN 2012bz: Bridging the Gap between Low- And High-Luminosity GRBs

Authors: S. Schulze, D. Malesani, A. Cucchiara, N. R. Tanvir, T. Krühler, A. de Ugarte Postigo, G. Leloudas, J. Lyman, D. Bersier, K. Wiersema, D. A. Perley, P. Schady, J. Gorosabel, J. P. Anderson, A. J. Castro-Tirado, S. B. Cenko, A. De Cia, L. E. Ellerbroek, J. P. U. Fynbo, J. Greiner, J. Hjorth, D. A. Kann, L. Kaper, S. Klose, A. J. Levan , et al. (40 additional authors not shown)

Abstract: At low redshift, a handful of gamma-ray bursts (GRBs) have been discovered with peak luminosities ($L_{\rm iso} < 10^{48.5}~\rm{erg\,s}^{-1}$) substantially lower than the average of the more distant ones ($L_{\rm iso} > 10^{49.5}~\rm{erg\,s}^{-1}$). The properties of several low-luminosity (low-$L$) GRBs indicate that they can be due to shock break-out, as opposed to the emission from ultrarelati… ▽ More At low redshift, a handful of gamma-ray bursts (GRBs) have been discovered with peak luminosities ($L_{\rm iso} < 10^{48.5}~\rm{erg\,s}^{-1}$) substantially lower than the average of the more distant ones ($L_{\rm iso} > 10^{49.5}~\rm{erg\,s}^{-1}$). The properties of several low-luminosity (low-$L$) GRBs indicate that they can be due to shock break-out, as opposed to the emission from ultrarelativistic jets. Owing to this, it is highly debated how both populations are connected, and whether there is a continuum between them. The burst at redshift $z=0.283$ from 2012 April 22 is one of the very few examples of intermediate-$L$ GRBs with a $γ$-ray luminosity of $L\sim10^{48.9}~\rm{erg\,s}^{-1}$ that have been detected up to now. Together with the robust detection of its accompanying supernova SN 2012bz, it has the potential to answer important questions on the origin of low- and high-$L$ GRBs and the GRB-SN connection. We carried out a spectroscopy campaign using medium- and low-resolution spectrographs at 6--10-m class telescopes, covering the time span of 37.3 days, and a multi-wavelength imaging campaign from radio to X-ray energies over a duration of $\sim270$ days. Furthermore, we used a tuneable filter centred at H$α$ to map star formation in the host galaxy and the surrounding galaxies. We used these data to extract and model the properties of different radiation components and incorporate spectral-energy-distribution fitting techniques to extract the properties of the host galaxy. Modelling the light curve and spectral energy distribution from the radio to the X-rays revealed the blast-wave to expand with an initial Lorentz factor of $Γ_0\sim60$, low for a high-$L$ GRB, and that the afterglow had an exceptional low peak luminosity-density of $\lesssim2\times10^{30}~\rm{erg\,s}^{-1}\,\rm{Hz}^{-1}$ in the sub-mm. [Abridged] △ Less

Submitted 15 January, 2014; originally announced January 2014.

Comments: 30 pages, 17 figures, 9 tables; abstract is abridged; images are shown at reduced resolution; comments are welcome

Afterglow rebrightenings as a signature of a long-lasting central engine activity? The emblematic case of GRB 100814A

Authors: M. Nardini, J. Elliott, R. Filgas, P. Schady, J. Greiner, T. Krühler, S. Klose, P. Afonso, D. A. Kann, A. Nicuesa Guelbenzu, F. Olivares E., A. Rau, A. Rossi, V. Sudilovsky, S. Schmidl

Abstract: In the past few years the number of well-sampled optical to NIR light curves of long Gamma-Ray Bursts (GRBs) has greatly increased particularly due to simultaneous multi-band imagers such as GROND. Combining these densely sampled ground-based data sets with the Swift UVOT and XRT space observations unveils a much more complex afterglow evolution than what was predicted by the most commonly invoked… ▽ More In the past few years the number of well-sampled optical to NIR light curves of long Gamma-Ray Bursts (GRBs) has greatly increased particularly due to simultaneous multi-band imagers such as GROND. Combining these densely sampled ground-based data sets with the Swift UVOT and XRT space observations unveils a much more complex afterglow evolution than what was predicted by the most commonly invoked theoretical models. GRB 100814A represents a remarkable example of these interesting well-sampled events, showing a prominent late-time rebrightening in the optical to NIR bands and a complex spectral evolution. This represents a unique laboratory to test the different afterglow emission models. Here we study the nature of the complex afterglow emission of GRB 100814A in the framework of different theoretical models. Moreover, we compare the late-time chromatic rebrightening with those observed in other well-sampled long GRBs. We analysed the optical and NIR observations obtained with the seven-channel Gamma-Ray burst Optical and Near-infrared Detector at the 2.2 m MPG/ESO telescope together with the X-ray and UV data detected by the instruments onboard the Swift observatory. The broad-band afterglow evolution, achieved by constructing multi-instrument light curves and spectral energy distributions, will be discussed in the framework of different theoretical models. We find that the standard models that describe the broad-band afterglow emission within the external shock scenario fail to describe the complex evolution of GRB 100814A, and therefore more complex scenarios must be invoked. [abridged] △ Less

Submitted 4 December, 2013; originally announced December 2013.

Comments: 11 pages, 7 figures, 2 tables; Astronomy & Astrophysics, in press

NuSTAR detection of the blazar B2 1023+25 at redshift 5.3

Authors: T. Sbarrato, G. Tagliaferri, G. Ghisellini, M. Perri, S. Puccetti, M. Balokovic, M. Nardini, D. Stern, S. E. Boggs, W. N. Brandt, F. E. Chirstensen, P. Giommi, J. Greiner, C. J. Hailey, F. Harrison, T. Hovatta, G. M. Madejski, A. Rau, P. Schady, V. Sudilovsky, C. M. Urry, W. W. Zhang

Abstract: B2 1023+25 is an extremely radio-loud quasar at z=5.3 which was first identified as a likely high-redshift blazar candidate in the SDSS+FIRST quasar catalog. Here we use the Nuclear Spectroscopic Telescope Array (NuSTAR) to investigate its non-thermal jet emission, whose high-energy component we detected in the hard X-ray energy band. The X-ray flux is ~5.5x10^(-14) erg cm^(-2)s^(-1) (5-10keV) and… ▽ More B2 1023+25 is an extremely radio-loud quasar at z=5.3 which was first identified as a likely high-redshift blazar candidate in the SDSS+FIRST quasar catalog. Here we use the Nuclear Spectroscopic Telescope Array (NuSTAR) to investigate its non-thermal jet emission, whose high-energy component we detected in the hard X-ray energy band. The X-ray flux is ~5.5x10^(-14) erg cm^(-2)s^(-1) (5-10keV) and the photon spectral index is Gamma_X=1.3-1.6. Modeling the full spectral energy distribution, we find that the jet is oriented close to the line of sight, with a viewing angle of ~3deg, and has significant Doppler boosting, with a large bulk Lorentz factor ~13, which confirms the identification of B2 1023+25 as a blazar. B2 1023+25 is the first object at redshift larger than 5 detected by NuSTAR, demonstrating the ability of NuSTAR to investigate the early X-ray Universe and to study extremely active supermassive black holes located at very high redshift. △ Less

Submitted 12 September, 2013; originally announced September 2013.

Comments: 10 pages, 3 figures, 3 tables. Accepted for publication in ApJ

The metallicity and dust content of a redshift 5 gamma-ray burst host galaxy

Authors: M. Sparre, O. E. Hartoog, T. Krühler, J. P. U. Fynbo, D. J. Watson, K. Wiersema, V. D'Elia, T. Zafar, P. M. J. Afonso, S. Covino, A. de Ugarte Postigo, H. Flores, P. Goldoni, J. Greiner, J. Hjorth, P. Jakobsson, L. Kaper, S. Klose, A. J. Levan, D. Malesani, B. Milvang-Jensen, M. Nardini, S. Piranomonte, J. Sollerman, R. Sánchez-Ramírez , et al. (4 additional authors not shown)

Abstract: Observations of the afterglows of long gamma-ray bursts (GRBs) allow the study of star-forming galaxies across most of cosmic history. Here we present observations of GRB 111008A from which we can measure metallicity, chemical abundance patterns, dust-to-metals ratio and extinction of the GRB host galaxy at z=5.0. The host absorption system is a damped Lyman-alpha absorber (DLA) with a very large… ▽ More Observations of the afterglows of long gamma-ray bursts (GRBs) allow the study of star-forming galaxies across most of cosmic history. Here we present observations of GRB 111008A from which we can measure metallicity, chemical abundance patterns, dust-to-metals ratio and extinction of the GRB host galaxy at z=5.0. The host absorption system is a damped Lyman-alpha absorber (DLA) with a very large neutral hydrogen column density of log N(HI)/cm^(-2) = 22.30 +/- 0.06, and a metallicity of [S/H]= -1.70 +/- 0.10. It is the highest redshift GRB with such a precise metallicity measurement. The presence of fine-structure lines confirms the z=5.0 system as the GRB host galaxy, and makes this the highest redshift where Fe II fine-structure lines have been detected. The afterglow is mildly reddened with A_V = 0.11 +/- 0.04 mag, and the host galaxy has a dust-to-metals ratio which is consistent with being equal to or lower than typical values in the Local Group. △ Less

Submitted 12 March, 2014; v1 submitted 11 September, 2013; originally announced September 2013.

Comments: Accepted for publication in ApJ

Journal ref: 2014 ApJ 785 150

The low-extinction afterglow in the solar-metallicity host galaxy of GRB 110918A

Authors: J. Elliott, T. Krühler, J. Greiner, S. Savaglio, F. Olivares E., A. Rau, A. de Ugarte Postigo, R. Sánchez-Ramírez, K. Wiersema, P. Schady, D. A. Kann, R. Filgas, M. Nardini, E. Berger, D. Fox, J. Gorosabel, S. Klose, A. Levan, A. Nicuesa Guelbenzu, A. Rossi, S. Schmidl, V. Sudilovsky, N. R. Tanvir, C. C. Thöne

Abstract: Galaxies selected through long gamma-ray bursts (GRBs) could be of fundamental importance when mapping the star formation history out to the highest redshifts. Before using them as efficient tools in the early Universe, however, the environmental factors that govern the formation of GRBs need to be understood. Metallicity is theoretically thought to be a fundamental driver in GRB explosions and en… ▽ More Galaxies selected through long gamma-ray bursts (GRBs) could be of fundamental importance when mapping the star formation history out to the highest redshifts. Before using them as efficient tools in the early Universe, however, the environmental factors that govern the formation of GRBs need to be understood. Metallicity is theoretically thought to be a fundamental driver in GRB explosions and energetics, but is still, even after more than a decade of extensive studies, not fully understood. This is largely related to two phenomena: a dust-extinction bias, that prevented high-mass and thus likely high-metallicity GRB hosts to be detected in the first place, and a lack of efficient instrumentation, that limited spectroscopic studies including metallicity measurements to the low-redshift end of the GRB host population. The subject of this work is the very energetic GRB 110918A, for which we measure one of the largest host-integrated metallicities, ever, and the highest stellar mass for z<1.9. This presents one of the very few robust metallicity measurements of GRB hosts at z~1, and establishes that GRB hosts at z~1 can also be very metal rich. It conclusively rules out a metallicity cut-off in GRB host galaxies and argues against an anti-correlation between metallicity and energy release in GRBs. △ Less

Submitted 26 August, 2013; originally announced August 2013.

Comments: 8 pages, 4 figures, 7th Huntsville Gamma-ray Burst Symposium, GRB 2013: paper 2 in eConf Proceedings C1304143

The low-extinction afterglow in the solar-metallicity host galaxy of gamma-ray burst 110918A

Authors: J. Elliott, T. Krühler, J. Greiner, S. Savaglio, F. Olivares E., A. Rau, A. de Ugarte Postigo, R. Sánchez-Ramírez, K. Wiersema, P. Schady, D. A. Kann, R. Filgas, M. Nardini, E. Berger, D. Fox, J. Gorosabel, S. Klose, A. Levan, A. Nicuesa Guelbenzu, A. Rossi, S. Schmidl, V. Sudilovsky, N. R. Tanvir, C. C. Thöne

Abstract: Metallicity is theoretically thought to be a fundamental driver in gamma-ray burst (GRB) explosions and energetics, but is still, even after more than a decade of extensive studies, not fully understood. This is largely related to two phenomena: a dust-extinction bias, that prevented high-mass and thus likely high-metallicity GRB hosts to be detected in the first place, and a lack of efficient ins… ▽ More Metallicity is theoretically thought to be a fundamental driver in gamma-ray burst (GRB) explosions and energetics, but is still, even after more than a decade of extensive studies, not fully understood. This is largely related to two phenomena: a dust-extinction bias, that prevented high-mass and thus likely high-metallicity GRB hosts to be detected in the first place, and a lack of efficient instrumentation, that limited spectroscopic studies including metallicity measurements to the low-redshift end of the GRB host population. The subject of this work is the very energetic GRB 110918A, for which we measure a redshift of z=0.984. GRB 110918A gave rise to a luminous afterglow with an intrinsic spectral slope of b=0.70, which probed a sight-line with little extinction (A_V=0.16 mag) typical of the established distributions of afterglow properties. Photometric and spectroscopic follow-up observations of the galaxy hosting GRB 110918A, including optical/NIR photometry with GROND and spectroscopy with VLT/X-shooter, however, reveal an all but average GRB host in comparison to the z~1 galaxies selected through similar afterglows to date. It has a large spatial extent with a half-light radius of ~10 kpc, the highest stellar mass for z<1.9 (log(M_*/M_sol) = 10.68+-0.16), and an Halpha-based star formation rate of 41 M_sol/yr. We measure a gas-phase extinction of ~1.8 mag through the Balmer decrement and one of the largest host-integrated metallicities ever of around solar (12 + log(O/H) = 8.93+/-0.13). This presents one of the very few robust metallicity measurements of GRB hosts at z~1, and establishes that GRB hosts at z~1 can also be very metal rich. It conclusively rules out a metallicity cut-off in GRB host galaxies and argues against an anti-correlation between metallicity and energy release in GRBs. △ Less

Submitted 4 June, 2013; originally announced June 2013.

Comments: 13 pages, 12 figures, 8 tables, accepted for publication in Astronomy & Astrophysics

Molecular Hydrogen in the Damped Lyman-alpha System towards GRB 120815A at z=2.36

Authors: T. Krühler, C. Ledoux, J. P. U. Fynbo, P. M. Vreeswijk, S. Schmidl, D. Malesani, L. Christensen, A. De Cia, J. Hjorth, P. Jakobsson, D. A. Kann, L. Kaper, S. D. Vergani, P. M. J. Afonso, S. Covino, A. de Ugarte Postigo, V. D'Elia, R. Filgas, P. Goldoni, J. Greiner, O. E. Hartoog, B. Milvang-Jensen, M. Nardini, S. Piranomonte, A. Rossi , et al. (10 additional authors not shown)

Abstract: We present the discovery of molecular hydrogen (H_2), including the presence of vibrationally-excited H_2^* in the optical spectrum of the afterglow of GRB 120815A at z=2.36 obtained with X-shooter at the VLT. Simultaneous photometric broad-band data from GROND and X-ray observations by Swift/XRT place further constraints on the amount and nature of dust along the sightline. The galactic environme… ▽ More We present the discovery of molecular hydrogen (H_2), including the presence of vibrationally-excited H_2^* in the optical spectrum of the afterglow of GRB 120815A at z=2.36 obtained with X-shooter at the VLT. Simultaneous photometric broad-band data from GROND and X-ray observations by Swift/XRT place further constraints on the amount and nature of dust along the sightline. The galactic environment of GRB 120815A is characterized by a strong DLA with log(N(H)/cm^-2) = 21.95 +/- 0.10, prominent H_2 absorption in the Lyman-Werner bands (log(N(H_2)/cm^-2) = 20.53 +/- 0.13) and thus a molecular gas fraction log f(H_2)=-1.14 +/- 0.15. The distance d between the absorbing neutral gas and GRB 120815A is constrained via photo-excitation modeling of fine-structure and meta-stable transitions of FeII and NiII to d = 0.5 +/- 0.1 kpc. The DLA metallicity ([Zn/H] = -1.15 +/- 0.12), visual extinction (A_V < 0.15 mag) and dust depletion ([Zn/Fe] = 1.01 +/- 0.10) are intermediate between the values of well-studied, H_2-deficient GRB-DLAs observed at high spectral resolution, and the approximately solar metallicity, highly-obscured and H_2-rich GRB 080607 sightline. With respect to N(H), metallicity, as well as dust-extinction and depletion, GRB 120815A is fairly representative of the average properties of GRB-DLAs. This demonstrates that molecular hydrogen is present in at least a fraction of the more typical GRB-DLAs, and H_2 and H_2^* are probably more wide-spread among GRB-selected systems than the few examples of previous detections would suggest. △ Less

Submitted 26 June, 2013; v1 submitted 25 April, 2013; originally announced April 2013.

Comments: 21 pages, 14 figures, accepted for publication in A&A

Journal ref: Astronomy & Astrophysics 557 (2013), A18

The unusual afterglow of the Gamma-Ray Burst 100621A

Authors: J. Greiner, T. Krühler, M. Nardini, R. Filgas, A. Moin, C. de Breuck, F. Montenegro-Montes, A. Lundgren, S. Klose, P. M. J. Afonso, F. Bertoldi, J. Elliott, D. A. Kann, F. Knust, K. Menten, A. Nicuesa Guelbenzu, F. Olivares E., A. Rau, A. Rossi, P. Schady, S. Schmidl, G. Siringo, L. Spezzi, V. Sudilovsky, S. J. Tingay , et al. (5 additional authors not shown)

Abstract: In order to constrain the broad-band spectral energy distribution of the afterglow of GRB 100621A, dedicated observations were performed in the optical/near-infrared with the 7-channel "Gamma-Ray Burst Optical and Near-infrared Detector" (GROND) at the 2.2m MPG/ESO telescope, in the sub-millimeter band with the large bolometer array LABOCA at APEX, and at radio frequencies with ATCA. Utilizing als… ▽ More In order to constrain the broad-band spectral energy distribution of the afterglow of GRB 100621A, dedicated observations were performed in the optical/near-infrared with the 7-channel "Gamma-Ray Burst Optical and Near-infrared Detector" (GROND) at the 2.2m MPG/ESO telescope, in the sub-millimeter band with the large bolometer array LABOCA at APEX, and at radio frequencies with ATCA. Utilizing also Swift X-ray observations, we attempt an interpretation of the observational data within the fireball scenario. The afterglow of GRB 100621A shows a very complex temporal as well as spectral evolution. We identify three different emission components, the most spectacular one causing a sudden intensity jump about one hour after the prompt emission. The spectrum of this component is much steeper than the canonical afterglow. We interpret this component using the prescription of Vlasis et al. (2011) for a two-shell collision after the first shell has been decelerated by the circumburst medium. We use the fireball scenario to derive constraints on the microphysical parameters of the first shell. Long-term energy injection into a narrow jet seems to provide an adequate description. Another noteworthy result is the large ($A_V$ = 3.6 mag) line-of-sight host extinction of the afterglow in an otherwise extremely blue host galaxy. △ Less

Submitted 22 April, 2013; originally announced April 2013.

Comments: 14 pages, 11 figs; acc. in A&A