Information Processing & Management最新文献_第4页

Instance and prototype contrastive learning for multi-view 3D model retrieval and classification 多视图三维模型检索与分类的实例与原型对比学习

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-28 DOI: 10.1016/j.ipm.2026.104655

Yaqian Zhou , Zhenghao Fang , Zhibin Gu , Song Yang , Yan Wang

Unsupervised 3D model retrieval and classification have obtained much attention due to widespread applications. However, existing methods focus only on global representations while ignoring local saliency learning, leading to redundant distraction and insufficient complementarity. In addition, they neglect intra- and inter-class contextual relevance during representation learning, leading to inaccurate embedding space partitioning and missing representative prototypes. To address these challenges, we present Instance and Prototype Contrastive Learning (IPCL), an unsupervised dual-network framework that simultaneously captures view-level local features and model-level semantic information. Specifically, we treat each view as an instance and employ inter-instance contrastive learning to extract discriminative local salient features, mitigating redundancy and enhancing cross-view complementarity. For global semantic modeling, we establish class prototypes for 3D models and propagate the semantic information to global features via a prototype-aware contrastive loss, strengthening class-level discriminability. Innovatively, we employ a bottom-up adaptive clustering algorithm called voting clustering, which mines deeper semantic correlations to refine prototype selection and embedding space structure. Comprehensive evaluations demonstrate the superiority of IPCL, e.g., IPCL outperforms most unsupervised methods, achieving classification accuracy improvements by 0.1% to 18.1% on ModelNet40 and 0.1% to 16.7% on ShapeNet55. IPCL achieves average retrieval gains of 15.2% in NN on ModelNet40 and 16.7% in mAP under the micro setting on ShapeNet55.

无监督三维模型检索与分类由于其广泛的应用而受到了广泛的关注。然而，现有的方法只关注全局表示，而忽略了局部显著性学习，导致冗余分心和互补性不足。此外，它们在表征学习过程中忽略了类内和类间的上下文相关性，导致嵌入空间划分不准确，缺失表征原型。为了解决这些挑战，我们提出了实例和原型对比学习（IPCL），这是一种无监督的双网络框架，可以同时捕获视图级局部特征和模型级语义信息。具体而言，我们将每个视图视为一个实例，并使用实例间对比学习来提取判别性的局部显著特征，减少冗余并增强跨视图互补性。对于全局语义建模，我们建立了三维模型的类原型，并通过原型感知的对比损失将语义信息传播到全局特征，增强了类级别的可判别性。创新之处是，我们采用了一种自下而上的自适应聚类算法，称为投票聚类，该算法挖掘更深层次的语义相关性来优化原型选择和嵌入空间结构。综合评价证明了IPCL的优越性，例如，IPCL优于大多数无监督方法，在ModelNet40上实现了0.1%至18.1%的分类精度提高，在ShapeNet55上实现了0.1%至16.7%的分类精度提高。在ShapeNet55的微设置下，IPCL在ModelNet40上的平均检索增益为15.2%，在mAP上的平均检索增益为16.7%。

{"title":"Instance and prototype contrastive learning for multi-view 3D model retrieval and classification","authors":"Yaqian Zhou , Zhenghao Fang , Zhibin Gu , Song Yang , Yan Wang","doi":"10.1016/j.ipm.2026.104655","DOIUrl":"10.1016/j.ipm.2026.104655","url":null,"abstract":"<div><div>Unsupervised 3D model retrieval and classification have obtained much attention due to widespread applications. However, existing methods focus only on global representations while ignoring local saliency learning, leading to redundant distraction and insufficient complementarity. In addition, they neglect intra- and inter-class contextual relevance during representation learning, leading to inaccurate embedding space partitioning and missing representative prototypes. To address these challenges, we present Instance and Prototype Contrastive Learning (IPCL), an unsupervised dual-network framework that simultaneously captures view-level local features and model-level semantic information. Specifically, we treat each view as an instance and employ inter-instance contrastive learning to extract discriminative local salient features, mitigating redundancy and enhancing cross-view complementarity. For global semantic modeling, we establish class prototypes for 3D models and propagate the semantic information to global features via a prototype-aware contrastive loss, strengthening class-level discriminability. Innovatively, we employ a bottom-up adaptive clustering algorithm called voting clustering, which mines deeper semantic correlations to refine prototype selection and embedding space structure. Comprehensive evaluations demonstrate the superiority of IPCL, e.g., IPCL outperforms most unsupervised methods, achieving classification accuracy improvements by 0.1% to 18.1% on ModelNet40 and 0.1% to 16.7% on ShapeNet55. IPCL achieves average retrieval gains of 15.2% in NN on ModelNet40 and 16.7% in mAP under the micro setting on ShapeNet55.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104655"},"PeriodicalIF":6.9,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081582","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A blockchain-based digital evidence management system: Integrating forensic procedures and multi-party authorization 基于区块链的数字证据管理系统：整合取证程序和多方授权

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-27 DOI: 10.1016/j.ipm.2026.104654

Yunji Park, Doowon Jeong

Current blockchain-based digital evidence systems provide strong technical integrity but fail to adequately address the procedural legitimacy required for court admissibility, frequently omitting judicial authorization workflows, differentiated handling of voluntary versus compulsory evidence, and transparent destruction protocols. To address these gaps, we propose B-DEMS, a blockchain-based digital evidence management system that integrates the full evidence lifecycle–from registration to court-authorized destruction–while encoding jurisdiction-specific legal requirements across South Korea, the United States, the European Union, and China. B-DEMS implements multi-party authorization, conditional decryption, and transaction-based disposal to ensure auditability and procedural compliance. Experimental evaluation across 1950 workflow executions demonstrated that B-DEMS achieved a maximum throughput of 10,890 TPS, representing 51–219% improvement over state-of-the-art systems, while maintaining stable scalability with latency increasing only 2.7-fold under a 5-fold peer expansion. Security analysis confirmed a 0% attack success rate across 300 adversarial attempts, and cross-border cooperation scenarios exhibited consistent adherence to jurisdiction-specific approval workflows. By aligning evidentiary procedures with a scalable blockchain architecture, B-DEMS provides a technically robust and procedurally compliant foundation for practical deployment in multi-agency and international investigative environments.

目前基于区块链的数字证据系统提供了强大的技术完整性，但未能充分解决法院可采性所需的程序合法性，经常忽略司法授权工作流程，区分自愿与强制证据的处理，以及透明的销毁协议。为了解决这些差距，我们提出了b - dem，这是一种基于区块链的数字证据管理系统，集成了从登记到法院授权销毁的完整证据生命周期，同时对韩国、美国、欧盟和中国的特定司法管辖区的法律要求进行编码。B-DEMS实现了多方授权、条件解密和基于事务的处理，以确保可审计性和程序遵从性。对1950个工作流执行的实验评估表明，b - dem实现了10,890 TPS的最大吞吐量，比最先进的系统提高了51-219%，同时保持稳定的可扩展性，延迟在5倍的对等扩展下仅增加2.7倍。安全分析证实，在300次对抗性攻击中，攻击成功率为0%，跨境合作场景显示出对特定管辖权审批工作流程的一致遵守。通过将证据程序与可扩展的区块链架构相结合，b - dem为在多机构和国际调查环境中实际部署提供了技术上强大和程序上兼容的基础。

{"title":"A blockchain-based digital evidence management system: Integrating forensic procedures and multi-party authorization","authors":"Yunji Park, Doowon Jeong","doi":"10.1016/j.ipm.2026.104654","DOIUrl":"10.1016/j.ipm.2026.104654","url":null,"abstract":"<div><div>Current blockchain-based digital evidence systems provide strong technical integrity but fail to adequately address the procedural legitimacy required for court admissibility, frequently omitting judicial authorization workflows, differentiated handling of voluntary versus compulsory evidence, and transparent destruction protocols. To address these gaps, we propose B-DEMS, a blockchain-based digital evidence management system that integrates the full evidence lifecycle–from registration to court-authorized destruction–while encoding jurisdiction-specific legal requirements across South Korea, the United States, the European Union, and China. B-DEMS implements multi-party authorization, conditional decryption, and transaction-based disposal to ensure auditability and procedural compliance. Experimental evaluation across 1950 workflow executions demonstrated that B-DEMS achieved a maximum throughput of 10,890 TPS, representing 51–219% improvement over state-of-the-art systems, while maintaining stable scalability with latency increasing only 2.7-fold under a 5-fold peer expansion. Security analysis confirmed a 0% attack success rate across 300 adversarial attempts, and cross-border cooperation scenarios exhibited consistent adherence to jurisdiction-specific approval workflows. By aligning evidentiary procedures with a scalable blockchain architecture, B-DEMS provides a technically robust and procedurally compliant foundation for practical deployment in multi-agency and international investigative environments.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104654"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SRCR: Faithful structured reasoning with curriculum reinforcement learning for explainable question answering SRCR：忠实的结构化推理与课程强化学习，用于可解释的问题回答

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-27 DOI: 10.1016/j.ipm.2026.104653

Yue Fan , Hu Zhang , Ru Li , Guangjun Zhang , Yujie Wang , Hongye Tan , Yuanlong Wang , Xiaoli Li , Jiye Liang

Existing explainable question answering methods based on structured reasoning lack effective modeling of logical dependencies between steps and underutilize the potential of intermediate conclusions in structured reasoning. To address these challenges, we propose SRCR, a faithful Structured Reasoning method based on Curriculum Reinforcement learning. Specifically, we propose an easy-to-difficult reverse structured curriculum that gradually slides the initial state of reasoning from end to beginning, which fully captures the complex dependencies of multi-step reasoning. Moreover, we treat fact selection and deductive generation as a unified process and construct a faithfulness reward function to mine faithful reasoning steps during the model learning and exploring phases. Experimental results on the structured reasoning datasets EntailmentBank and STREET demonstrate that SRCR achieves state-of-the-art performance in factual accuracy and intermediate conclusion correctness, surpassing previous methods by 8.0% and 2.0%, respectively. Moreover, SRCR also improves answer accuracy by 2.6% to 8.3%, and extensive analysis shows that SRCR can generate more faithful structured explanations.

现有的基于结构化推理的可解释问答方法缺乏对步骤间逻辑依赖关系的有效建模，未能充分利用结构化推理中中间结论的潜力。为了解决这些挑战，我们提出了SRCR，一种基于课程强化学习的忠实结构化推理方法。具体来说，我们提出了一个易难的反向结构化课程，逐步将推理的初始状态从头到尾滑动，充分捕捉了多步骤推理的复杂依赖关系。此外，我们将事实选择和演绎生成视为一个统一的过程，并构建忠实度奖励函数来挖掘模型学习和探索阶段的忠实推理步骤。在结构化推理数据集EntailmentBank和STREET上的实验结果表明，SRCR在事实准确性和中间结论正确性方面达到了最先进的性能，分别比以前的方法提高了8.0%和2.0%。此外，SRCR还将答案准确率提高了2.6%至8.3%，广泛的分析表明，SRCR可以生成更忠实的结构化解释。

{"title":"SRCR: Faithful structured reasoning with curriculum reinforcement learning for explainable question answering","authors":"Yue Fan , Hu Zhang , Ru Li , Guangjun Zhang , Yujie Wang , Hongye Tan , Yuanlong Wang , Xiaoli Li , Jiye Liang","doi":"10.1016/j.ipm.2026.104653","DOIUrl":"10.1016/j.ipm.2026.104653","url":null,"abstract":"<div><div>Existing explainable question answering methods based on structured reasoning lack effective modeling of logical dependencies between steps and underutilize the potential of intermediate conclusions in structured reasoning. To address these challenges, we propose SRCR, a faithful <strong>S</strong>tructured <strong>R</strong>easoning method based on <strong>C</strong>urriculum <strong>R</strong>einforcement learning. Specifically, we propose an easy-to-difficult reverse structured curriculum that gradually slides the initial state of reasoning from end to beginning, which fully captures the complex dependencies of multi-step reasoning. Moreover, we treat fact selection and deductive generation as a unified process and construct a faithfulness reward function to mine faithful reasoning steps during the model learning and exploring phases. Experimental results on the structured reasoning datasets EntailmentBank and STREET demonstrate that SRCR achieves state-of-the-art performance in factual accuracy and intermediate conclusion correctness, surpassing previous methods by 8.0% and 2.0%, respectively. Moreover, SRCR also improves answer accuracy by 2.6% to 8.3%, and extensive analysis shows that SRCR can generate more faithful structured explanations.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104653"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081585","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SPECTRA-Net: Spatiotemporal edge-preserving contextual reinforcement architecture for adaptive crowd behavior recognition 光谱网络：用于自适应人群行为识别的时空边缘保持上下文强化结构

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-27 DOI: 10.1016/j.ipm.2026.104647

Min Zhu , Dengyin Zhang

The study presents the HERA-Net (Hierarchical Edge-aware Reinforcement Architecture) framework, which combines Hierarchical Motion Saliency (HMS) and Deep Reinforcement Learning (DRL) for adaptive crowd behavior recognition. The UCSD Ped2 dataset, comprising 32 surveillance clips (240 × 360 px), showed that HERA-Net improved generalization performance by 20 %, resilience to occlusion by 18 %, and recognition accuracy by 12–15 % compared to state-of-the-art models. In dynamic crowd situations, the HMS module hierarchically mixes local and global motion cues to maintain edge boundaries, while the DRL policy adaptively enhances recognition. A PPO-based DRL enables real-time adaptive behavior detection, and a unique edge-aware loss function ensures exact motion boundaries. Experimental results demonstrate that HERA-Net successfully balances precision and adaptability, making it a dependable, real-time system for intelligent surveillance, anomaly identification, and crowd monitoring.

该研究提出了HERA-Net（分层边缘感知强化架构）框架，该框架将分层运动显著性（HMS）和深度强化学习（DRL）相结合，用于自适应人群行为识别。UCSD Ped2数据集包含32个监控片段（240 × 360像素），表明与最先进的模型相比，HERA-Net的泛化性能提高了20%，遮挡复原力提高了18%，识别精度提高了12 - 15%。在动态人群情况下，HMS模块分层混合局部和全局运动线索以保持边缘边界，而DRL策略自适应增强识别。基于ppo的DRL实现实时自适应行为检测，独特的边缘感知损失功能确保精确的运动边界。实验结果表明，HERA-Net成功地平衡了精度和适应性，使其成为一种可靠、实时的智能监控、异常识别和人群监控系统。

引用次数: 0

Diffusion -driven group anomaly detection in spatiotemporal trajectories: Robust masked sequence imputation for enhanced pattern discovery 时空轨迹中扩散驱动的群体异常检测：增强模式发现的鲁棒掩码序列输入

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-27 DOI: 10.1016/j.ipm.2026.104650

Qi Ouyang , Hongchang Chen , Shuxin Liu , Ran Li , Kai Wang , Yingle Li

Detecting group anomalies in spatiotemporal trajectories is vital for smart city security but remains challenging due to noise and complex dynamics. Existing diffusion models often suffer from “conditional contamination,” where observed anomalies inadvertently guide reconstruction, masking the very deviations they aim to detect. To address this, we propose Diffusion-Driven Group Anomaly Detection (DGAD). Specifically, we introduce an Interleaved Window Masking Strategy that segments data to enforce mutual supervision and reveal latent patterns. We pair this with an Unconditional Imputation Mechanism that conditions generation on forward noise instead of partial observations. This prevents anomaly leakage and significantly widens the divergence between normal and abnormal behaviors. Furthermore, a Denoising Weighted Voting module aggregates outputs across diffusion steps to mitigate uncertainty and enhance stability. Extensive experiments on synthetic and real-world datasets show that DGAD consistently outperforms state-of-the-art methods, improving F1-score by 1.1% on average and 1.8% in high-noise conditions. Code and case studies are available at [https://github.com/oyq-star/DAGD-main].

探测时空轨迹中的群体异常对于智慧城市安全至关重要，但由于噪声和复杂的动态，仍然具有挑战性。现有的扩散模型经常受到“条件污染”的影响，即观察到的异常在无意中指导了重建，掩盖了它们旨在检测的偏差。为了解决这个问题，我们提出了扩散驱动的群体异常检测（DGAD）。具体来说，我们引入了一种交错窗口掩蔽策略，该策略将数据分割以加强相互监督并揭示潜在模式。我们将其与无条件归算机制配对，该机制以前向噪声而不是部分观测值为条件生成。这样可以防止异常泄漏，并大大扩大了正常和异常行为之间的差异。此外，一个去噪加权投票模块聚合了跨扩散步骤的输出，以减轻不确定性并增强稳定性。在合成数据集和真实数据集上进行的大量实验表明，DGAD始终优于最先进的方法，平均将f1分数提高1.1%，在高噪声条件下提高1.8%。代码和案例研究可在[https://github.com/oyq-star/DAGD-main]]获得。

{"title":"Diffusion -driven group anomaly detection in spatiotemporal trajectories: Robust masked sequence imputation for enhanced pattern discovery","authors":"Qi Ouyang , Hongchang Chen , Shuxin Liu , Ran Li , Kai Wang , Yingle Li","doi":"10.1016/j.ipm.2026.104650","DOIUrl":"10.1016/j.ipm.2026.104650","url":null,"abstract":"<div><div>Detecting group anomalies in spatiotemporal trajectories is vital for smart city security but remains challenging due to noise and complex dynamics. Existing diffusion models often suffer from “conditional contamination,” where observed anomalies inadvertently guide reconstruction, masking the very deviations they aim to detect. To address this, we propose Diffusion-Driven Group Anomaly Detection (DGAD). Specifically, we introduce an Interleaved Window Masking Strategy that segments data to enforce mutual supervision and reveal latent patterns. We pair this with an Unconditional Imputation Mechanism that conditions generation on forward noise instead of partial observations. This prevents anomaly leakage and significantly widens the divergence between normal and abnormal behaviors. Furthermore, a Denoising Weighted Voting module aggregates outputs across diffusion steps to mitigate uncertainty and enhance stability. Extensive experiments on synthetic and real-world datasets show that DGAD consistently outperforms state-of-the-art methods, improving F1-score by 1.1% on average and 1.8% in high-noise conditions. Code and case studies are available at [<span><span>https://github.com/oyq-star/DAGD-main</span><svg><path></path></svg></span>].</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104650"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081586","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

GAI-HIQ: Developing a health information quality assessment indicator system for generative artificial intelligence 基于生成式人工智能的健康信息质量评估指标体系研究

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-27 DOI: 10.1016/j.ipm.2026.104651

Jia Yi , Fei Du , Yuchen Nie , Wencui Liang , Xiaoyu Zhou , Jinjuan Chen , Guangying Li , Mimi Liu , Yalan Lv , Wenlong Zhao , Xiaorong Hou

The rapid adoption of generative artificial intelligence (GAI) in healthcare has raised concerns about the quality of its outputs, yet existing health information assessment tools are not designed for AI-generated content. This study developed the GAI-HIQ assessment scale. Using Latent Dirichlet Allocation on 341 relevant publications, we initially identified three core dimensions and 15 secondary indicators for evaluating health information quality. A two-round Delphi consultation with 20 experts (100 % response rate) refined the framework to three core dimensions and 13 secondary indicators, achieving a statistically significant level of expert consensus (Kendall’s W = 0.288, P < 0.05). The analytic hierarchy process was then applied to calculate indicator weights, ensuring structured prioritization of quality dimensions (all judgment matrices yielded CR values below 0.1). The GAI-HIQ provides a consensus-based framework for evaluating health information generated by AI, offering practical tools for developers to optimize algorithms, for healthcare institutions to regulate applications, and for patients to assess information reliability.

生成式人工智能（GAI）在医疗保健领域的迅速采用引发了对其产出质量的担忧，但现有的健康信息评估工具并不是为人工智能生成的内容设计的。本研究编制了GAI-HIQ评估量表。通过对341份相关出版物的潜在狄利克雷分配，我们初步确定了评估健康信息质量的三个核心维度和15个次要指标。通过20位专家（100%回复率）的两轮德尔菲咨询，将框架细化为三个核心维度和13个次要指标，专家共识达到了具有统计学意义的水平（Kendall 's W = 0.288, P < 0.05）。然后应用层次分析法计算指标权重，确保质量维度的结构化优先级（所有判断矩阵的CR值均低于0.1）。AI- hiq为评估人工智能生成的健康信息提供了基于共识的框架，为开发人员优化算法、卫生保健机构规范应用程序和患者评估信息可靠性提供了实用工具。

{"title":"GAI-HIQ: Developing a health information quality assessment indicator system for generative artificial intelligence","authors":"Jia Yi , Fei Du , Yuchen Nie , Wencui Liang , Xiaoyu Zhou , Jinjuan Chen , Guangying Li , Mimi Liu , Yalan Lv , Wenlong Zhao , Xiaorong Hou","doi":"10.1016/j.ipm.2026.104651","DOIUrl":"10.1016/j.ipm.2026.104651","url":null,"abstract":"<div><div>The rapid adoption of generative artificial intelligence (GAI) in healthcare has raised concerns about the quality of its outputs, yet existing health information assessment tools are not designed for AI-generated content. This study developed the GAI-HIQ assessment scale. Using Latent Dirichlet Allocation on 341 relevant publications, we initially identified three core dimensions and 15 secondary indicators for evaluating health information quality. A two-round Delphi consultation with 20 experts (100 % response rate) refined the framework to three core dimensions and 13 secondary indicators, achieving a statistically significant level of expert consensus (Kendall’s <em>W</em> = 0.288, <em>P</em> < 0.05). The analytic hierarchy process was then applied to calculate indicator weights, ensuring structured prioritization of quality dimensions (all judgment matrices yielded CR values below 0.1). The GAI-HIQ provides a consensus-based framework for evaluating health information generated by AI, offering practical tools for developers to optimize algorithms, for healthcare institutions to regulate applications, and for patients to assess information reliability.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104651"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081584","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Positive and negative neighbor dual-flexible nonnegative representation method for image classification 图像分类的正邻域和负邻域双柔性非负表示方法

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-27 DOI: 10.1016/j.ipm.2026.104645

Ziqi Li , Zhilin Chen , Tingting Guo , Yonghong Zhang , Xiaoning Song , Tianyang Xu

Nonnegative representation-based classification (NRC) enhances the discriminability of samples to some extent through nonnegativity constraints. However, this approach often fails to fully exploit local feature structures and class-discriminative information, which limits its overall classification performance. To overcome this limitation, we introduce the positive and negative neighbor dual-flexible nonnegative representation (PN²DFNR) classifier. This method integrates both the positive and negative neighbor sets of the query sample to effectively capture location information. Specifically, the positive neighbor set enforces spatial consistency, ensuring that the estimated representation of the query sample remains well-aligned with its neighboring samples. In contrast, the negative neighbor set introduces inverse constraints to suppress inter-class interference. Furthermore, a flexibility factor is incorporated to formulate a weighted flexible constraint strategy, which enhances the representational capacity of the correct class while adaptively attenuating the contribution of incorrect classes. To evaluate PN²DFNR’s performance, extensive experiments are conducted on facial recognition, handwritten digit classification, and large-scale datasets with diverse feature characteristics. The results demonstrate that PN²DFNR achieves superior classification performance, with the maximum improvement reaching approximately 2%. The model code will be available on the author’s homepage (https://github.com/li-zi-qi/PN2DFNR).

基于非负表示的分类（NRC）通过非负性约束在一定程度上增强了样本的可判别性。然而，这种方法往往不能充分利用局部特征结构和分类判别信息，从而限制了其整体分类性能。为了克服这一限制，我们引入了正邻域和负邻域双柔性非负表示（PN2DFNR）分类器。该方法结合查询样本的正邻集和负邻集，有效地捕获位置信息。具体来说，正邻居集加强了空间一致性，确保查询样本的估计表示与其相邻样本保持良好对齐。相反，负邻居集引入逆约束来抑制类间干扰。在此基础上，引入柔性因子制定了加权柔性约束策略，增强了正确类的表示能力，同时自适应地降低了错误类的贡献。为了评估PN2DFNR的性能，我们在人脸识别、手写数字分类和具有不同特征特征的大规模数据集上进行了大量的实验。结果表明，PN2DFNR具有优异的分类性能，最大改进幅度约为2%。模型代码可以在作者的主页上找到（https://github.com/li-zi-qi/PN2DFNR）。

{"title":"Positive and negative neighbor dual-flexible nonnegative representation method for image classification","authors":"Ziqi Li , Zhilin Chen , Tingting Guo , Yonghong Zhang , Xiaoning Song , Tianyang Xu","doi":"10.1016/j.ipm.2026.104645","DOIUrl":"10.1016/j.ipm.2026.104645","url":null,"abstract":"<div><div>Nonnegative representation-based classification (NRC) enhances the discriminability of samples to some extent through nonnegativity constraints. However, this approach often fails to fully exploit local feature structures and class-discriminative information, which limits its overall classification performance. To overcome this limitation, we introduce the positive and negative neighbor dual-flexible nonnegative representation (PN<sup>2</sup>DFNR) classifier. This method integrates both the positive and negative neighbor sets of the query sample to effectively capture location information. Specifically, the positive neighbor set enforces spatial consistency, ensuring that the estimated representation of the query sample remains well-aligned with its neighboring samples. In contrast, the negative neighbor set introduces inverse constraints to suppress inter-class interference. Furthermore, a flexibility factor is incorporated to formulate a weighted flexible constraint strategy, which enhances the representational capacity of the correct class while adaptively attenuating the contribution of incorrect classes. To evaluate PN<sup>2</sup>DFNR’s performance, extensive experiments are conducted on facial recognition, handwritten digit classification, and large-scale datasets with diverse feature characteristics. The results demonstrate that PN<sup>2</sup>DFNR achieves superior classification performance, with the maximum improvement reaching approximately 2%. The model code will be available on the author’s homepage (<span><span>https://github.com/li-zi-qi/PN2DFNR</span><svg><path></path></svg></span>).</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104645"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045206","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

PERStance: Personality-guided enhanced multimodal stance detection 持久性：个性导向的增强多模态姿态检测

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-23 DOI: 10.1016/j.ipm.2025.104593

Guoqi Geng, Qianyi Zhan, Heng-Yang Lu

Stance detection aims to identify users’ attitudes toward specific targets in social media, playing a crucial role in information processing and public opinion management. However, existing research on stance detection often overlooks the potential influence of user personality traits on stance expression. In view of the shortcomings of existing stance detection methods, this paper studies the impact of personality traits on stance judgment. To this end, we propose PERStance, a personality-guided enhanced multimodal zero-shot stance detection method. Specifically, PERStance uses a Large Language Model (LLM) to infer users’ personality traits from a multi-dimensional perspective, thereby more accurately understanding users’ potential stances on specific targets. To mitigate issues such as LLM hallucinations and reasoning confusion, we incorporate the Chain-of-Thought framework in the stance detection stage and optimize its reasoning path. Experimental results on multiple multimodal stance detection datasets show that the PERStance method proposed in this paper achieves the best performance in stance detection, with an average increase of 23.88% in the Macro-F1 score. Ablation experiments verify the effectiveness of each module in this method. The source code of our proposed framework is released at https://github.com/jncsnlp/PERStance.

姿态检测旨在识别用户在社交媒体中对特定目标的态度，在信息处理和舆论管理中起着至关重要的作用。然而，现有的姿态检测研究往往忽略了用户人格特征对姿态表达的潜在影响。针对现有姿态检测方法的不足，本文研究了人格特征对姿态判断的影响。为此，我们提出了一种人格引导的增强多模态零射击姿态检测方法PERStance。具体而言，PERStance使用LLM （Large Language Model）从多维度角度推断用户的人格特征，从而更准确地了解用户对特定目标的潜在立场。为了减轻LLM幻觉和推理混乱等问题，我们将思维链框架纳入姿态检测阶段，并优化其推理路径。在多个多模态姿态检测数据集上的实验结果表明，本文提出的PERStance方法在姿态检测中表现最佳，其Macro-F1得分平均提高23.88%。烧蚀实验验证了该方法中各个模块的有效性。我们提出的框架的源代码发布在https://github.com/jncsnlp/PERStance。

{"title":"PERStance: Personality-guided enhanced multimodal stance detection","authors":"Guoqi Geng, Qianyi Zhan, Heng-Yang Lu","doi":"10.1016/j.ipm.2025.104593","DOIUrl":"10.1016/j.ipm.2025.104593","url":null,"abstract":"<div><div>Stance detection aims to identify users’ attitudes toward specific targets in social media, playing a crucial role in information processing and public opinion management. However, existing research on stance detection often overlooks the potential influence of user personality traits on stance expression. In view of the shortcomings of existing stance detection methods, this paper studies the impact of personality traits on stance judgment. To this end, we propose PERStance, a personality-guided enhanced multimodal zero-shot stance detection method. Specifically, PERStance uses a Large Language Model (LLM) to infer users’ personality traits from a multi-dimensional perspective, thereby more accurately understanding users’ potential stances on specific targets. To mitigate issues such as LLM hallucinations and reasoning confusion, we incorporate the Chain-of-Thought framework in the stance detection stage and optimize its reasoning path. Experimental results on multiple multimodal stance detection datasets show that the PERStance method proposed in this paper achieves the best performance in stance detection, with an average increase of <strong>23.88%</strong> in the Macro-F1 score. Ablation experiments verify the effectiveness of each module in this method. The source code of our proposed framework is released at <span><span>https://github.com/jncsnlp/PERStance</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 4","pages":"Article 104593"},"PeriodicalIF":6.9,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146023312","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Validating generative agent-Based modeling in social media simulations through the lens of the friendship paradox 通过友谊悖论的镜头验证社交媒体模拟中基于生成代理的建模

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-23 DOI: 10.1016/j.ipm.2026.104636

Gian Marco Orlando , Valerio La Gatta , Diego Russo , Vincenzo Moscato

Generative Agent-Based Modeling (GABM) is an emerging simulation paradigm that integrates the reasoning capabilities of Large Language Models (LLMs) with traditional Agent-Based Modeling to replicate complex social behaviors, including user interactions on social media platforms. While GABM has been employed to study localized phenomena on social media, such as opinion formation and information propagation, its capacity to capture global network-level phenomena remains underexplored. In this paper, we address this gap by investigating whether GABM-based social media simulations exhibit the Friendship Paradox (FP) – a counterintuitive network phenomenon where individuals, on average, have fewer friends than their friends. We design a GABM-based framework for social media simulation, featuring generative agents that emulate real users by incorporating distinct personalities, interests, and behaviors. Leveraging three real-world Twitter datasets centered on the US 2020 Election, UK Brexit, and the QAnon conspiracy, we demonstrate that the FP and its generalized forms emerge in GABM-based simulations. Consistent with real-world social media, we observe a hierarchical structure where generative agents preferentially connect with others exhibiting superior attributes, such as greater activity or influence, without being instructed with any behavioral rules. Furthermore, our analysis reveals that infrequent connections with highly connected agents primarily drive the Friendship Paradox, mirroring established patterns in real-world networks. Overall, our findings validate the ability of GABM to replicate global social media phenomena, highlighting its potential as a robust framework for modeling and analyzing complex social behaviors at scale.

基于生成代理的建模（GABM）是一种新兴的仿真范式，它将大型语言模型（llm）的推理能力与传统的基于代理的建模相结合，以复制复杂的社会行为，包括社交媒体平台上的用户交互。虽然GABM已被用于研究社交媒体上的局部现象，如意见形成和信息传播，但其捕捉全球网络层面现象的能力仍未得到充分探索。在本文中，我们通过调查基于gabm的社交媒体模拟是否表现出友谊悖论（FP）来解决这一差距。FP是一种反直觉的网络现象，平均而言，个人的朋友比他们的朋友少。我们设计了一个基于gabm的社交媒体模拟框架，其特点是生成代理通过结合不同的个性、兴趣和行为来模拟真实用户。利用以2020年美国大选、英国脱欧和QAnon阴谋为中心的三个真实世界的Twitter数据集，我们证明了FP及其广义形式出现在基于gabm的模拟中。与现实世界的社交媒体一致，我们观察到一种层次结构，在这种结构中，生成智能体优先与表现出更优越属性的其他智能体连接，比如更活跃或更有影响力，而不受任何行为规则的指导。此外，我们的分析表明，与高度连接的代理之间的不频繁连接主要驱动了友谊悖论，反映了现实世界网络中的既定模式。总的来说，我们的研究结果验证了GABM复制全球社交媒体现象的能力，突出了它作为大规模建模和分析复杂社会行为的强大框架的潜力。

{"title":"Validating generative agent-Based modeling in social media simulations through the lens of the friendship paradox","authors":"Gian Marco Orlando , Valerio La Gatta , Diego Russo , Vincenzo Moscato","doi":"10.1016/j.ipm.2026.104636","DOIUrl":"10.1016/j.ipm.2026.104636","url":null,"abstract":"<div><div>Generative Agent-Based Modeling (GABM) is an emerging simulation paradigm that integrates the reasoning capabilities of Large Language Models (LLMs) with traditional Agent-Based Modeling to replicate complex social behaviors, including user interactions on social media platforms. While GABM has been employed to study localized phenomena on social media, such as opinion formation and information propagation, its capacity to capture global network-level phenomena remains underexplored. In this paper, we address this gap by investigating whether GABM-based social media simulations exhibit the Friendship Paradox (FP) – a counterintuitive network phenomenon where individuals, on average, have fewer friends than their friends. We design a GABM-based framework for social media simulation, featuring <em>generative agents</em> that emulate real users by incorporating distinct personalities, interests, and behaviors. Leveraging three real-world Twitter datasets centered on the US 2020 Election, UK Brexit, and the QAnon conspiracy, we demonstrate that the FP and its generalized forms emerge in GABM-based simulations. Consistent with real-world social media, we observe a hierarchical structure where <em>generative agents</em> preferentially connect with others exhibiting superior attributes, such as greater activity or influence, without being instructed with any behavioral rules. Furthermore, our analysis reveals that infrequent connections with highly connected agents primarily drive the Friendship Paradox, mirroring established patterns in real-world networks. Overall, our findings validate the ability of GABM to replicate global social media phenomena, highlighting its potential as a robust framework for modeling and analyzing complex social behaviors at scale.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 4","pages":"Article 104636"},"PeriodicalIF":6.9,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146023310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Conversational generative retrieval with contextual denoising 会话生成检索与上下文去噪

IF 6.9 1区管理学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Processing & Management

Pub Date : 2026-01-23 DOI: 10.1016/j.ipm.2026.104642

Wenhao Zhang , Yuhan Wang , Yi Bai , Mengqi Zhang , Zhaochun Ren , Zhumin Chen , Pengjie Ren

Conversational search needs an understanding of the interaction history between users and systems to retrieve information that aligns with the user’s current query. A prominent challenge in this task lies in addressing irrelevant context, which can obscure the user’s true intent and hinder retrieval accuracy. Existing methods for conversational search rely on the dense retrieval architecture, which encodes the context into fixed-length vectors. However, these approaches are constrained by the inherent limitations of the bi-encoder architecture, making it difficult to mitigate interference from irrelevant contexts effectively. In this paper, we propose Conversational Generative Retrieval for Context Denoising (CGR4CD), a novel framework that leverages a sequence-to-sequence generative retrieval architecture with cross-attention layers to dynamically capture key information during decoding. Our method introduces innovative proposition-based docids and a two-stage training strategy, combining generation and rank losses for enhanced denoising and ranking. Extensive evaluations on three conversational search benchmarks (TopiOCQA, QReCC, and OR-QuAC) demonstrate that CGR4CD achieves state-of-the-art performance, outperforming baselines in 8 out of 9 evaluation metrics. On TopiOCQA, which is characterized by topic switching and high noise, CGR4CD achieves an average improvement of 8.2% over baselines of similar size and surpasses the LLM-based Chatretriever by 4.4%, highlighting its superior context denoising capability. Our approach also maintains parameter efficiency, achieving these results with a T5-large backbone compared to larger LLM-based retrievers. Detailed analysis further validates its superior capability in contextual denoising, particularly in noisy scenarios.

会话搜索需要了解用户和系统之间的交互历史，以便检索与用户当前查询一致的信息。该任务的一个突出挑战在于处理不相关的上下文，这可能会模糊用户的真实意图并影响检索的准确性。现有的会话搜索方法依赖于密集检索架构，该架构将上下文编码为固定长度的向量。然而，这些方法受到双编码器架构固有局限性的限制，使得难以有效地减轻来自不相关上下文的干扰。在本文中，我们提出了上下文去噪的会话生成检索（CGR4CD），这是一个新颖的框架，它利用具有跨注意层的序列到序列生成检索架构来动态捕获解码过程中的关键信息。我们的方法引入了创新的基于命题的docids和两阶段训练策略，结合了生成和秩损失来增强去噪和排序。对三个会话搜索基准（TopiOCQA、QReCC和OR-QuAC）的广泛评估表明，CGR4CD达到了最先进的性能，在9个评估指标中有8个优于基线。在主题切换和高噪声特征的TopiOCQA上，CGR4CD比相似大小的基线平均提高了8.2%，比基于llm的Chatretriever高出4.4%，突出了其优越的上下文去噪能力。我们的方法还保持了参数效率，与基于llm的大型检索器相比，使用t5大的主干获得了这些结果。详细的分析进一步验证了其在上下文去噪方面的优越能力，特别是在嘈杂的场景下。

{"title":"Conversational generative retrieval with contextual denoising","authors":"Wenhao Zhang , Yuhan Wang , Yi Bai , Mengqi Zhang , Zhaochun Ren , Zhumin Chen , Pengjie Ren","doi":"10.1016/j.ipm.2026.104642","DOIUrl":"10.1016/j.ipm.2026.104642","url":null,"abstract":"<div><div>Conversational search needs an understanding of the interaction history between users and systems to retrieve information that aligns with the user’s current query. A prominent challenge in this task lies in addressing irrelevant context, which can obscure the user’s true intent and hinder retrieval accuracy. Existing methods for conversational search rely on the dense retrieval architecture, which encodes the context into fixed-length vectors. However, these approaches are constrained by the inherent limitations of the bi-encoder architecture, making it difficult to mitigate interference from irrelevant contexts effectively. In this paper, we propose Conversational Generative Retrieval for Context Denoising (CGR4CD), a novel framework that leverages a sequence-to-sequence generative retrieval architecture with cross-attention layers to dynamically capture key information during decoding. Our method introduces innovative proposition-based docids and a two-stage training strategy, combining generation and rank losses for enhanced denoising and ranking. Extensive evaluations on three conversational search benchmarks (TopiOCQA, QReCC, and OR-QuAC) demonstrate that CGR4CD achieves state-of-the-art performance, outperforming baselines in 8 out of 9 evaluation metrics. On TopiOCQA, which is characterized by topic switching and high noise, CGR4CD achieves an average improvement of 8.2% over baselines of similar size and surpasses the LLM-based Chatretriever by 4.4%, highlighting its superior context denoising capability. Our approach also maintains parameter efficiency, achieving these results with a T5-large backbone compared to larger LLM-based retrievers. Detailed analysis further validates its superior capability in contextual denoising, particularly in noisy scenarios.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 4","pages":"Article 104642"},"PeriodicalIF":6.9,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146023311","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0