Pub Date : 2026-01-20DOI: 10.1016/j.inffus.2026.104164
Yunfei Guo
The proliferation of urban big data presents unprecedented opportunities for understanding cities, yet the analytical methods to harness this data are often fragmented and domain-specific. Existing predictive models in urban computing are typically highly specialized, creating analytical silos that inhibit knowledge transfer and are difficult to adapt across domains such as public safety, housing and transport. This paper confronts this critical gap by developing a generalizable, multimodal spatio-temporal deep learning framework engineered for both high predictive performance and interpretability, which is capable of mastering diverse urban prediction tasks without architectural modification. The hybrid architecture fuses a Multi-Head Graph Convolutional Network (GCN) for spatial diffusion, a Long Short-Term Memory (LSTM) network for temporal dynamics, and a learnable Gating Mechanism that weights the influence of spatial graph versus static external features. To validate this generalizability, the framework was tested on three distinct urban domains in London: crime forecasting, housing price estimation and transport network demand. The model outperformed traditional baselines (ARIMA, XGBoost) and state-of-the-art deep learning models (TabNet, TFT). Moreover, the framework moves beyond prediction to explanation by incorporating attention mechanisms and permutation feature importance analysis.
{"title":"Multimodal spatio-temporal fusion: A generalizable GCN-LSTM with attention framework for urban application","authors":"Yunfei Guo","doi":"10.1016/j.inffus.2026.104164","DOIUrl":"10.1016/j.inffus.2026.104164","url":null,"abstract":"<div><div>The proliferation of urban big data presents unprecedented opportunities for understanding cities, yet the analytical methods to harness this data are often fragmented and domain-specific. Existing predictive models in urban computing are typically highly specialized, creating analytical silos that inhibit knowledge transfer and are difficult to adapt across domains such as public safety, housing and transport. This paper confronts this critical gap by developing a generalizable, multimodal spatio-temporal deep learning framework engineered for both high predictive performance and interpretability, which is capable of mastering diverse urban prediction tasks without architectural modification. The hybrid architecture fuses a Multi-Head Graph Convolutional Network (GCN) for spatial diffusion, a Long Short-Term Memory (LSTM) network for temporal dynamics, and a learnable Gating Mechanism that weights the influence of spatial graph versus static external features. To validate this generalizability, the framework was tested on three distinct urban domains in London: crime forecasting, housing price estimation and transport network demand. The model outperformed traditional baselines (ARIMA, XGBoost) and state-of-the-art deep learning models (TabNet, TFT). Moreover, the framework moves beyond prediction to explanation by incorporating attention mechanisms and permutation feature importance analysis.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104164"},"PeriodicalIF":15.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014809","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-20DOI: 10.1016/j.inffus.2026.104165
Shengyingjie Liu , Jianxin Li , Qian Wan , Bo He , Zhijun Huang , Qing Li
Programming education is essential for equipping individuals with digital literacy skills and developing the problem-solving abilities necessary for success in the modern workforce. In online programming tutoring systems, knowledge tracing (KT) techniques are crucial for programming prediction, as they monitor user performance and model user cognition. However, both universal and programming-specific knowledge transfer methods depend on traditional state-driven paradigms that indirectly predict programming outcomes based on users’ knowledge states. It does not align with the core objective of programming prediction, which is to determine whether submitted code can solve the question. To address this, we present the code-driven feature fusion KT (CFKT), which integrates large language models (LLM) and encoders for both individualized and common code features. It consists of two modules: pass prediction and code prediction. The pass prediction module leverages LLM to incorporate semantic information from the question and code through embedding, extracting key features that determine code correctness through proxy tasks and effectively narrowing the solution space with vectorization. The code prediction module integrates user historical data and data from other users through feature fusion blocks, allowing for accurate predictions of submitted code and effectively mitigating the cold start problem. Experiments on multiple real-world public programming datasets demonstrate that CFKT significantly outperforms existing baseline methods.
{"title":"Code-driven programming prediction enhanced by LLM with a feature fusion approach","authors":"Shengyingjie Liu , Jianxin Li , Qian Wan , Bo He , Zhijun Huang , Qing Li","doi":"10.1016/j.inffus.2026.104165","DOIUrl":"10.1016/j.inffus.2026.104165","url":null,"abstract":"<div><div>Programming education is essential for equipping individuals with digital literacy skills and developing the problem-solving abilities necessary for success in the modern workforce. In online programming tutoring systems, knowledge tracing (KT) techniques are crucial for programming prediction, as they monitor user performance and model user cognition. However, both universal and programming-specific knowledge transfer methods depend on traditional state-driven paradigms that indirectly predict programming outcomes based on users’ knowledge states. It does not align with the core objective of programming prediction, which is to determine whether submitted code can solve the question. To address this, we present the code-driven feature fusion KT (CFKT), which integrates large language models (LLM) and encoders for both individualized and common code features. It consists of two modules: pass prediction and code prediction. The pass prediction module leverages LLM to incorporate semantic information from the question and code through embedding, extracting key features that determine code correctness through proxy tasks and effectively narrowing the solution space with vectorization. The code prediction module integrates user historical data and data from other users through feature fusion blocks, allowing for accurate predictions of submitted code and effectively mitigating the cold start problem. Experiments on multiple real-world public programming datasets demonstrate that CFKT significantly outperforms existing baseline methods.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104165"},"PeriodicalIF":15.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014543","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-20DOI: 10.1016/j.inffus.2026.104163
Kang Liu , Shunzhi Yang , Chang-Dong Wang , Yunwen Chen , Zhenhua Huang
The primary goal of graph knowledge distillation (GKD) is to transfer knowledge from a complex graph neural network (GNN) teacher to a smaller, yet more efficient GNN or multi-layer perceptron student. Although existing methods address network scalability, they rely on a frozen teacher that fails to explain how to derive results, thus limiting performance and hindering the improvement of a student. Therefore, we propose a novel GKD method, termed Dynamic Gradient Distillation (DGD), consisting of Generative Adversarial Imitation Learning (GAIL)-based Gradient Mapping and Two-Stage Gradient Fusion modules. The former builds the teacher’s learning process to understand knowledge by drawing on the principle of GAIL. The latter consists of attention fusion and weighted bias operations. Through the attentional fusion operation, it captures and fuses the responses of the teacher to change the gradient of the student at each layer. The fused gradients are then updated by combining them with the student’s backpropagated gradients using the weighted bias operation. DGD allows the student to inherit and extend the teacher’s learning process efficiently. Extensive experiments conducted with seven publicly available datasets show that DGD could significantly outperform some existing methods in node classification tasks. Our code and data are released at https://github.com/KangL-G/Dynamic-Gradient-Distillation.
{"title":"A novel knowledge distillation method for graph neural networks with gradient mapping and fusion","authors":"Kang Liu , Shunzhi Yang , Chang-Dong Wang , Yunwen Chen , Zhenhua Huang","doi":"10.1016/j.inffus.2026.104163","DOIUrl":"10.1016/j.inffus.2026.104163","url":null,"abstract":"<div><div>The primary goal of graph knowledge distillation (GKD) is to transfer knowledge from a complex graph neural network (GNN) teacher to a smaller, yet more efficient GNN or multi-layer perceptron student. Although existing methods address network scalability, they rely on a frozen teacher that fails to explain how to derive results, thus limiting performance and hindering the improvement of a student. Therefore, we propose a novel GKD method, termed Dynamic Gradient Distillation (DGD), consisting of Generative Adversarial Imitation Learning (GAIL)-based Gradient Mapping and Two-Stage Gradient Fusion modules. The former builds the teacher’s learning process to understand knowledge by drawing on the principle of GAIL. The latter consists of attention fusion and weighted bias operations. Through the attentional fusion operation, it captures and fuses the responses of the teacher to change the gradient of the student at each layer. The fused gradients are then updated by combining them with the student’s backpropagated gradients using the weighted bias operation. DGD allows the student to inherit and extend the teacher’s learning process efficiently. Extensive experiments conducted with seven publicly available datasets show that DGD could significantly outperform some existing methods in node classification tasks. Our code and data are released at <span><span>https://github.com/KangL-G/Dynamic-Gradient-Distillation</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104163"},"PeriodicalIF":15.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146006526","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-20DOI: 10.1016/j.inffus.2026.104166
Yongcheng Li , Lingcong Cai , Ying Lu , Xiao Han , Ma Li , Wenxing Lai , Xiangzhong Zhang , Xiaomao Fan
Blood cell identification is critical for hematological analysis as it aids physicians in diagnosing various blood-related diseases. In real-world scenarios, blood cell image datasets often present the issues of domain shift and data imbalance, posing challenges for accurate blood cell identification. To address these issues, we propose a novel blood cell classification method termed SADA via stain-aware domain alignment. The primary objective of this work is to mine domain-invariant features in the presence of domain shifts and data imbalance. To accomplish this objective, we propose a stain-based augmentation approach and a local alignment constraint to learn domain-invariant features. Furthermore, we propose a domain-invariant supervised contrastive learning strategy to capture discriminative features. We decouple the training process into two stages of domain-invariant feature learning and classification training, alleviating the problem of data imbalance. Experiment results on four public blood cell datasets and a private real dataset collected from the Third Affiliated Hospital of Sun Yat-sen University demonstrate that SADA can achieve a new state-of-the-art baseline, which is superior to the existing cutting-edge methods. The source code can be available at the URL (https://github.com/AnoK3111/SADA).
{"title":"Stain-aware domain alignment for imbalance blood cell classification","authors":"Yongcheng Li , Lingcong Cai , Ying Lu , Xiao Han , Ma Li , Wenxing Lai , Xiangzhong Zhang , Xiaomao Fan","doi":"10.1016/j.inffus.2026.104166","DOIUrl":"10.1016/j.inffus.2026.104166","url":null,"abstract":"<div><div>Blood cell identification is critical for hematological analysis as it aids physicians in diagnosing various blood-related diseases. In real-world scenarios, blood cell image datasets often present the issues of domain shift and data imbalance, posing challenges for accurate blood cell identification. To address these issues, we propose a novel blood cell classification method termed SADA via stain-aware domain alignment. The primary objective of this work is to mine domain-invariant features in the presence of domain shifts and data imbalance. To accomplish this objective, we propose a stain-based augmentation approach and a local alignment constraint to learn domain-invariant features. Furthermore, we propose a domain-invariant supervised contrastive learning strategy to capture discriminative features. We decouple the training process into two stages of domain-invariant feature learning and classification training, alleviating the problem of data imbalance. Experiment results on four public blood cell datasets and a private real dataset collected from the Third Affiliated Hospital of Sun Yat-sen University demonstrate that SADA can achieve a new state-of-the-art baseline, which is superior to the existing cutting-edge methods. The source code can be available at the URL (<span><span>https://github.com/AnoK3111/SADA</span><svg><path></path></svg></span>).</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104166"},"PeriodicalIF":15.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-19DOI: 10.1016/j.inffus.2026.104162
Wuzhen Shi , Wu Yang , Zhihao Wu , Yang Wen
Existing prior-guided image inpainting methods show state-of-the-art performance. But their prior extraction is computationally expensive and unstable in accuracy. Besides, most of them only focus on the structure guidance, which hardly facilitates the repair of realistic textures. Inspired by the fact that gradient maps are easy to extract and reflect both image structure and fine texture details, this paper proposes a gradient-guided network for image inpainting, which first uses the gradient context information and multi-level image compensation features to repair the gradient, and then uses the repaired gradient features to guide the generation of realistic image. A gradient-driven attention (GDA) module is introduced for efficient prior guidance. Additionally, a context validity-aware (CVA) module is proposed for progressively filling hole regions of images, which accurately utilizes both local and contextual information for image inpainting via validity-aware measurements. Furthermore, by artificially manipulating the generation of the gradient map, our gradient-guided image inpainting method enables user-guided image editing, which effectively increases the diversity of image generation and enhances the flexibility of image editing. Experiments on benchmark datasets show that the proposed method outperforms the state-of-the-art methods. Extensive ablation experiments are also conducted to demonstrate the effectiveness of each module.
{"title":"Validity-aware context modeling for gradient-guided image inpainting","authors":"Wuzhen Shi , Wu Yang , Zhihao Wu , Yang Wen","doi":"10.1016/j.inffus.2026.104162","DOIUrl":"10.1016/j.inffus.2026.104162","url":null,"abstract":"<div><div>Existing prior-guided image inpainting methods show state-of-the-art performance. But their prior extraction is computationally expensive and unstable in accuracy. Besides, most of them only focus on the structure guidance, which hardly facilitates the repair of realistic textures. Inspired by the fact that gradient maps are easy to extract and reflect both image structure and fine texture details, this paper proposes a gradient-guided network for image inpainting, which first uses the gradient context information and multi-level image compensation features to repair the gradient, and then uses the repaired gradient features to guide the generation of realistic image. A gradient-driven attention (GDA) module is introduced for efficient prior guidance. Additionally, a context validity-aware (CVA) module is proposed for progressively filling hole regions of images, which accurately utilizes both local and contextual information for image inpainting via validity-aware measurements. Furthermore, by artificially manipulating the generation of the gradient map, our gradient-guided image inpainting method enables user-guided image editing, which effectively increases the diversity of image generation and enhances the flexibility of image editing. Experiments on benchmark datasets show that the proposed method outperforms the state-of-the-art methods. Extensive ablation experiments are also conducted to demonstrate the effectiveness of each module.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104162"},"PeriodicalIF":15.5,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146000897","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-18DOI: 10.1016/j.inffus.2026.104124
Gabriel Oduori , Chaira Cocco , Payam Sajadi , Francesco Pilla
Data fusion (DF) addresses the challenge of integrating heterogeneous data sources to improve decision-making and inference. Although DF has been widely explored, no prior systematic review has specifically focused on its application to low-cost sensor (LCS) data in environmental monitoring. To address this gap, we conduct a systematic literature review (SLR) following the PRISMA framework, synthesising findings from 82 peer-reviewed articles. The review addresses three key questions: (1) What fusion methodologies are employed in conjunction with LCS data? (2) In what environmental contexts are these methods applied? (3) What are the methodological challenges and research gaps? Our analysis reveals that geostatistical and machine learning approaches dominate current practice, with air quality monitoring emerging as the primary application domain. Additionally, artificial intelligence (AI)-based methods are increasingly used to integrate spatial, temporal, and multimodal data. However, limitations persist in uncertainty quantification, validation standards, and the generalisability of fusion frameworks. This review provides a comprehensive synthesis of current techniques and outlines key directions for future research, including the development of robust, uncertainty-aware fusion methods and broader application to less-studied environmental variables.
{"title":"Data fusion for low-cost sensors: A systematic literature review","authors":"Gabriel Oduori , Chaira Cocco , Payam Sajadi , Francesco Pilla","doi":"10.1016/j.inffus.2026.104124","DOIUrl":"10.1016/j.inffus.2026.104124","url":null,"abstract":"<div><div>Data fusion (DF) addresses the challenge of integrating heterogeneous data sources to improve decision-making and inference. Although DF has been widely explored, no prior systematic review has specifically focused on its application to low-cost sensor (LCS) data in environmental monitoring. To address this gap, we conduct a systematic literature review (SLR) following the PRISMA framework, synthesising findings from 82 peer-reviewed articles. The review addresses three key questions: (1) What fusion methodologies are employed in conjunction with LCS data? (2) In what environmental contexts are these methods applied? (3) What are the methodological challenges and research gaps? Our analysis reveals that geostatistical and machine learning approaches dominate current practice, with air quality monitoring emerging as the primary application domain. Additionally, artificial intelligence (AI)-based methods are increasingly used to integrate spatial, temporal, and multimodal data. However, limitations persist in uncertainty quantification, validation standards, and the generalisability of fusion frameworks. This review provides a comprehensive synthesis of current techniques and outlines key directions for future research, including the development of robust, uncertainty-aware fusion methods and broader application to less-studied environmental variables.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104124"},"PeriodicalIF":15.5,"publicationDate":"2026-01-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145995206","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-18DOI: 10.1016/j.inffus.2026.104161
Adil Chakhtouna, Sara Sekkate, Abdellah Adib
Speech Emotion Recognition (SER) is a rapidly evolving research field that aims to enable machines to automatically identify human emotions from vocal signals. This systematic review presents a comprehensive and structured synthesis of the SER literature, focusing on eleven key research questions that span the theoretical foundations, signal processing pipeline, and methodological advancements in the field. Unlike prior surveys, this review unifies both foundational and state-of-the-art insights across all stages of the SER pipeline through a question-driven structure, offering a clear road-map for both new and experienced researchers in the SER community. We first explore the psychological and computational modeling of emotions, followed by a detailed examination of the different modalities for emotion expression, with a particular emphasis on speech. The review highlights the most widely used emotional speech databases, common pre-processing techniques, and the diverse set of handcrafted and learned features employed in SER. We compare traditional machine learning approaches with recent deep learning models, emphasizing their respective strengths, limitations, and application contexts. Special attention is given to the recent shift toward self-supervised learning (SSL) models such as Wav2Vec2 and HuBERT, which have redefined the state-of-the-art in speech-based representation learning. Special attention is given to evaluation metrics, benchmarking strategies, and real-world deployment challenges, including issues of speaker-independence and environmental variability. The review concludes by identifying key limitations across the literature and articulating future research directions necessary for developing reliable, scalable, and context-aware emotion-aware systems. Overall, this work serves as a central reference for advancing SER research and practical deployment in real-world environments.
{"title":"Speech emotion recognition: A systematic mega-review of techniques and pipelines","authors":"Adil Chakhtouna, Sara Sekkate, Abdellah Adib","doi":"10.1016/j.inffus.2026.104161","DOIUrl":"10.1016/j.inffus.2026.104161","url":null,"abstract":"<div><div>Speech Emotion Recognition (SER) is a rapidly evolving research field that aims to enable machines to automatically identify human emotions from vocal signals. This systematic review presents a comprehensive and structured synthesis of the SER literature, focusing on eleven key research questions that span the theoretical foundations, signal processing pipeline, and methodological advancements in the field. Unlike prior surveys, this review unifies both foundational and state-of-the-art insights across all stages of the SER pipeline through a question-driven structure, offering a clear road-map for both new and experienced researchers in the SER community. We first explore the psychological and computational modeling of emotions, followed by a detailed examination of the different modalities for emotion expression, with a particular emphasis on speech. The review highlights the most widely used emotional speech databases, common pre-processing techniques, and the diverse set of handcrafted and learned features employed in SER. We compare traditional machine learning approaches with recent deep learning models, emphasizing their respective strengths, limitations, and application contexts. Special attention is given to the recent shift toward self-supervised learning (SSL) models such as Wav2Vec2 and HuBERT, which have redefined the state-of-the-art in speech-based representation learning. Special attention is given to evaluation metrics, benchmarking strategies, and real-world deployment challenges, including issues of speaker-independence and environmental variability. The review concludes by identifying key limitations across the literature and articulating future research directions necessary for developing reliable, scalable, and context-aware emotion-aware systems. Overall, this work serves as a central reference for advancing SER research and practical deployment in real-world environments.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104161"},"PeriodicalIF":15.5,"publicationDate":"2026-01-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145995202","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-17DOI: 10.1016/j.inffus.2026.104157
Xunqi Zhou , Zhenqi Zhang , Zifeng Wu , Qianming Wang , Jing Teng , Jinlong Liu , Yongjie Zhai
In intelligent vehicle damage assessment, component recognition faces challenges such as significant intra-class variability and minimal inter-class differences, which hinder detection, as well as occlusions and ambiguous boundaries, which complicate segmentation. We generalize these problems into three core aspects: inter-object relational modeling, semantic-detail information balancing, and occlusion-aware decoupling. To this end, we propose the Adaptive Regularized Topological Segmentation (ARTSeg) network, comprising three complementary modules: Inter-Class Graph Constraint (ICGC), Constrained Detail Feature Backtracking (CDFB), and Topological Decoupling Segmentation (TDS). Each module is purposefully designed, integrated in a progressive structure, and synergistically reinforces the others to enhance overall performance. Specifically, ICGC clusters intra-class features and establishes implicit topological constraints among categories during feature extraction, enabling the model to better capture inter-class relationships and improve detection representation. Subsequently, CDFB evaluates the impact of channel-wise feature information within each candidate region on segmentation accuracy and computational cost, dynamically selecting appropriate feature resolutions for individual instances while balancing the demands of detection and segmentation tasks. Finally, TDS introduces topological associations between occluded and occluding regions at the feature level and decouples them at the task level, explicitly modeling generalized occlusion regions and enhancing segmentation performance. We quantitatively and qualitatively evaluate ARTSeg on a 59-category vehicle component dataset constructed for insurance damage assessment, achieving notable improvements in addressing the aforementioned problems. Experiments on two public datasets, DSMLR and Carparts, further validate the generalization capability of the proposed method. Results indicate that ARTSeg provides practical guidance for component recognition in intelligent vehicle damage assessment.
{"title":"An adaptive regularized topological segmentation network integrating inter-class relations and occlusion information for vehicle component recognition","authors":"Xunqi Zhou , Zhenqi Zhang , Zifeng Wu , Qianming Wang , Jing Teng , Jinlong Liu , Yongjie Zhai","doi":"10.1016/j.inffus.2026.104157","DOIUrl":"10.1016/j.inffus.2026.104157","url":null,"abstract":"<div><div>In intelligent vehicle damage assessment, component recognition faces challenges such as significant intra-class variability and minimal inter-class differences, which hinder detection, as well as occlusions and ambiguous boundaries, which complicate segmentation. We generalize these problems into three core aspects: inter-object relational modeling, semantic-detail information balancing, and occlusion-aware decoupling. To this end, we propose the Adaptive Regularized Topological Segmentation (ARTSeg) network, comprising three complementary modules: Inter-Class Graph Constraint (ICGC), Constrained Detail Feature Backtracking (CDFB), and Topological Decoupling Segmentation (TDS). Each module is purposefully designed, integrated in a progressive structure, and synergistically reinforces the others to enhance overall performance. Specifically, ICGC clusters intra-class features and establishes implicit topological constraints among categories during feature extraction, enabling the model to better capture inter-class relationships and improve detection representation. Subsequently, CDFB evaluates the impact of channel-wise feature information within each candidate region on segmentation accuracy and computational cost, dynamically selecting appropriate feature resolutions for individual instances while balancing the demands of detection and segmentation tasks. Finally, TDS introduces topological associations between occluded and occluding regions at the feature level and decouples them at the task level, explicitly modeling generalized occlusion regions and enhancing segmentation performance. We quantitatively and qualitatively evaluate ARTSeg on a 59-category vehicle component dataset constructed for insurance damage assessment, achieving notable improvements in addressing the aforementioned problems. Experiments on two public datasets, DSMLR and Carparts, further validate the generalization capability of the proposed method. Results indicate that ARTSeg provides practical guidance for component recognition in intelligent vehicle damage assessment.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104157"},"PeriodicalIF":15.5,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145995203","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-17DOI: 10.1016/j.inffus.2026.104160
Minghan Li , Ercong Nie , Huiping Huang , Xinxuan Lv , Guodong Zhou
Large Language Models (LLMs) show strong potential for query expansion (QE), but their effectiveness is highly sensitive to prompt design. This paper investigates whether exploiting the system-user prompt distinction in chat-based LLMs improves QE, and how multiple expansions should be combined. We propose Dual-Layer Prompt Ensembles, which pair a behavioural system prompt with varied user prompts to generate diverse expansions, and aggregate their BM25-ranked lists using lightweight SU-RankFusion schemes. Experiments on six heterogeneous datasets show that dual-layer prompting consistently outperforms strong single-prompt baselines. For example, on Touche-2020 a dual-layer configuration improves nDCG@10 from 0.4177 (QE-CoT) to 0.4696, and SU-RankFusion further raises it to 0.4797. On Robust04 and DBPedia, SU-RankFusion improves nDCG@10 over BM25 by 24.7% and 25.5%, respectively, with similar gains on NFCorpus, FiQA, and TREC-COVID. These results demonstrate that system-user prompt ensembles are effective for QE, and that simple fusion transforms prompt-level diversity into stable retrieval improvements.
{"title":"Dual-layer prompt ensembles: Leveraging system- and user-level instructions for robust LLM-based query expansion and rank fusion","authors":"Minghan Li , Ercong Nie , Huiping Huang , Xinxuan Lv , Guodong Zhou","doi":"10.1016/j.inffus.2026.104160","DOIUrl":"10.1016/j.inffus.2026.104160","url":null,"abstract":"<div><div>Large Language Models (LLMs) show strong potential for query expansion (QE), but their effectiveness is highly sensitive to prompt design. This paper investigates whether exploiting the system-user prompt distinction in chat-based LLMs improves QE, and how multiple expansions should be combined. We propose Dual-Layer Prompt Ensembles, which pair a behavioural system prompt with varied user prompts to generate diverse expansions, and aggregate their BM25-ranked lists using lightweight SU-RankFusion schemes. Experiments on six heterogeneous datasets show that dual-layer prompting consistently outperforms strong single-prompt baselines. For example, on Touche-2020 a dual-layer configuration improves nDCG@10 from 0.4177 (QE-CoT) to 0.4696, and SU-RankFusion further raises it to 0.4797. On Robust04 and DBPedia, SU-RankFusion improves nDCG@10 over BM25 by 24.7% and 25.5%, respectively, with similar gains on NFCorpus, FiQA, and TREC-COVID. These results demonstrate that system-user prompt ensembles are effective for QE, and that simple fusion transforms prompt-level diversity into stable retrieval improvements.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"131 ","pages":"Article 104160"},"PeriodicalIF":15.5,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145995213","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-17DOI: 10.1016/j.inffus.2026.104152
Juan Carlos González-Quesada , Ignacio Javier Pérez , Enrique Herrera-Viedma , Francisco Javier Cabrerizo
Granular computing is a framework encompassing tools, techniques, and theories that utilize information granules to address complex problems. Recently, it has become a popular area of study for managing uncertainty in group decision-making. Numerous models using granular computing have been developed to tackle issues such as incomplete information, consistency, and consensus in fuzzy group decision-making. However, existing granular-based approaches fail to consider two critical factors in managing consensus: (i) the individual’s willingness to engage and (ii) the necessity of mitigating bias in interpersonal interactions. To address these gaps, we propose a new granular consensus-reaching process inspired by the blockchain technology, which helps create trust among participants. Unlike most previous methods, our approach minimizes biased interactions among participants by using a communication structure based on blockchain and smart contracts. In this setup, participants’ identities, opinions, and decisions regarding the acceptance or rejection of received recommendations remain private from other peers. Additionally, our approach includes a trust-building mechanism, also based on blockchain, encouraging individuals to rethink and adjust their opinions. It differs from most previous trust-building methods by removing the requirement of opinion similarity and avoiding trust propagation. Instead, it builds trust among participants by allowing them to see how many peers have accepted suggested modifications. This enhances computational efficiency in creating trust and speeds up consensus. To demonstrate how effective our approach is, we provide a numerical example, along with a sensitivity analysis of its key assumptions and a discussion of its strengths and weaknesses. The results confirm that this new granular consensus-reaching process is valid, effective, and practical.
{"title":"A granular consensus-reaching process using blockchain-based mechanisms to foster trust relationships in fuzzy group decision-making","authors":"Juan Carlos González-Quesada , Ignacio Javier Pérez , Enrique Herrera-Viedma , Francisco Javier Cabrerizo","doi":"10.1016/j.inffus.2026.104152","DOIUrl":"10.1016/j.inffus.2026.104152","url":null,"abstract":"<div><div>Granular computing is a framework encompassing tools, techniques, and theories that utilize information granules to address complex problems. Recently, it has become a popular area of study for managing uncertainty in group decision-making. Numerous models using granular computing have been developed to tackle issues such as incomplete information, consistency, and consensus in fuzzy group decision-making. However, existing granular-based approaches fail to consider two critical factors in managing consensus: (i) the individual’s willingness to engage and (ii) the necessity of mitigating bias in interpersonal interactions. To address these gaps, we propose a new granular consensus-reaching process inspired by the blockchain technology, which helps create trust among participants. Unlike most previous methods, our approach minimizes biased interactions among participants by using a communication structure based on blockchain and smart contracts. In this setup, participants’ identities, opinions, and decisions regarding the acceptance or rejection of received recommendations remain private from other peers. Additionally, our approach includes a trust-building mechanism, also based on blockchain, encouraging individuals to rethink and adjust their opinions. It differs from most previous trust-building methods by removing the requirement of opinion similarity and avoiding trust propagation. Instead, it builds trust among participants by allowing them to see how many peers have accepted suggested modifications. This enhances computational efficiency in creating trust and speeds up consensus. To demonstrate how effective our approach is, we provide a numerical example, along with a sensitivity analysis of its key assumptions and a discussion of its strengths and weaknesses. The results confirm that this new granular consensus-reaching process is valid, effective, and practical.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104152"},"PeriodicalIF":15.5,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145995204","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}