AI Open最新文献 - Book学术

TRiSM for Agentic AI: A review of Trust, Risk, and Security Management in LLM-based Agentic Multi-Agent Systems 代理人工智能的TRiSM：基于法学硕士的代理多代理系统中的信任、风险和安全管理综述

IF 14.8

AI Open

Pub Date : 2026-01-01 Epub Date: 2026-03-02 DOI: 10.1016/j.aiopen.2026.02.006

Shaina Raza , Ranjan Sapkota , Manoj Karkee , Christos Emmanouilidis

Agentic AI systems, built upon large language models (LLMs) and deployed in multi-agent configurations, are redefining intelligence, autonomy, collaboration, and decision-making across enterprise and societal domains. This review presents a structured analysis of Trust, Risk, and Security Management (TRiSM) in the context of LLM-based Agentic Multi-Agent Systems (AMAS). We begin by examining the conceptual foundations of Agentic AI and highlight its architectural distinctions from traditional AI agents. We then adapt and extend the AI TRiSM framework for Agentic AI, structured around key pillars: Explainability, ModelOps, Security, Privacy and their Lifecycle Governance, each contextualized to the challenges of AMAS. A risk taxonomy is proposed to capture the unique threats and vulnerabilities of Agentic AI, ranging from coordination failures to prompt-based adversarial manipulation. To make coordination and tool use measurable in practice, we propose two metrics: the Component Synergy Score (CSS), which captures inter-agent enablement, and the Tool Utilization Efficacy (TUE), which evaluates whether tools are invoked correctly and efficiently. We further discuss strategies for improving explainability in Agentic AI, as well as approaches to enhancing security and privacy through encryption, adversarial robustness, and regulatory compliance. The review concludes with a research roadmap for the responsible development and deployment of Agentic AI, highlighting key directions to align emerging systems with TRiSM principles-ensuring safety, transparency, and accountability in their operation.

基于大型语言模型（llm）并部署在多智能体配置中的人工智能系统正在重新定义跨企业和社会领域的智能、自治、协作和决策。本文综述了基于法学硕士的代理多代理系统（AMAS）背景下的信任、风险和安全管理（TRiSM）的结构化分析。我们首先研究人工智能代理的概念基础，并强调其与传统人工智能代理在架构上的区别。然后，我们为人工智能调整和扩展了人工智能TRiSM框架，围绕关键支柱构建：可解释性、ModelOps、安全性、隐私及其生命周期治理，每个支柱都针对AMAS的挑战。提出了一种风险分类法来捕捉人工智能的独特威胁和漏洞，从协调失败到基于提示的对抗性操纵。为了使协调和工具使用在实践中可测量，我们提出了两个度量标准：组件协同得分（CSS），它捕获代理间的启用，以及工具利用效率（TUE），它评估工具是否被正确和有效地调用。我们进一步讨论了提高人工智能可解释性的策略，以及通过加密、对抗性鲁棒性和法规遵从性来增强安全性和隐私性的方法。报告最后提出了负责任开发和部署人工智能的研究路线图，强调了使新兴系统与TRiSM原则保持一致的关键方向，即确保其运行中的安全性、透明度和问责制。

{"title":"TRiSM for Agentic AI: A review of Trust, Risk, and Security Management in LLM-based Agentic Multi-Agent Systems","authors":"Shaina Raza , Ranjan Sapkota , Manoj Karkee , Christos Emmanouilidis","doi":"10.1016/j.aiopen.2026.02.006","DOIUrl":"10.1016/j.aiopen.2026.02.006","url":null,"abstract":"<div><div>Agentic AI systems, built upon large language models (LLMs) and deployed in multi-agent configurations, are redefining intelligence, autonomy, collaboration, and decision-making across enterprise and societal domains. This review presents a structured analysis of Trust, Risk, and Security Management (TRiSM) in the context of LLM-based Agentic Multi-Agent Systems (AMAS). We begin by examining the conceptual foundations of Agentic AI and highlight its architectural distinctions from traditional AI agents. We then adapt and extend the AI TRiSM framework for Agentic AI, structured around key pillars: <em>Explainability, ModelOps, Security, Privacy</em> and <em>their Lifecycle Governance</em>, each contextualized to the challenges of AMAS. A risk taxonomy is proposed to capture the unique threats and vulnerabilities of Agentic AI, ranging from coordination failures to prompt-based adversarial manipulation. To make coordination and tool use measurable in practice, we propose two metrics: the Component Synergy Score (CSS), which captures inter-agent enablement, and the Tool Utilization Efficacy (TUE), which evaluates whether tools are invoked correctly and efficiently. We further discuss strategies for improving explainability in Agentic AI, as well as approaches to enhancing security and privacy through encryption, adversarial robustness, and regulatory compliance. The review concludes with a research roadmap for the responsible development and deployment of Agentic AI, highlighting key directions to align emerging systems with TRiSM principles-ensuring safety, transparency, and accountability in their operation.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"7 ","pages":"Pages 71-95"},"PeriodicalIF":14.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147404227","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

GHOST 2.0: Generative high-fidelity one shot transfer of heads GHOST 2.0：生成高保真的一次性头部转移

IF 14.8

AI Open

Pub Date : 2026-01-01 Epub Date: 2026-02-14 DOI: 10.1016/j.aiopen.2026.02.003

Alexander Groshev , Anastasiia Iashchenko , Pavel Paramonov , Denis Dimitrov , Andrey Kuznetsov

While the task of face swapping has recently gained attention in the research community, a related problem of head swapping remains largely unexplored. In addition to skin color transfer, head swap poses extra challenges, such as the need to preserve structural information of the whole head during synthesis and inpaint gaps between swapped head and background. In this paper, we address these concerns with GHOST 2.0, which consists of two problem-specific modules. First, we introduce enhanced Aligner model for head reenactment, which preserves identity information at multiple scales and is robust to extreme pose variations. Secondly, we use a Blender module that seamlessly integrates the reenacted head into the target background by transferring skin color and inpainting mismatched regions. Both modules outperform the baselines on the corresponding tasks, allowing to achieve state-of-the-art results in head swapping. We also tackle complex cases, such as large difference in hair styles of source and target.

虽然换脸的任务最近引起了研究界的关注，但换头的相关问题在很大程度上仍未被探索。除了皮肤颜色的转移，头部交换还带来了额外的挑战，例如在合成过程中需要保留整个头部的结构信息，以及交换头部与背景之间的油漆间隙。在本文中，我们用GHOST 2.0解决了这些问题，它由两个问题特定的模块组成。首先，我们引入了用于头部再现的增强Aligner模型，该模型在多尺度上保留了身份信息，并且对极端姿态变化具有鲁棒性。其次，我们使用一个Blender模块，通过转移皮肤颜色和油漆不匹配的区域，无缝地将再现的头部集成到目标背景中。这两个模块在相应的任务上都优于基线，从而实现了最先进的头部交换结果。我们也处理复杂的情况，例如来源和目标的发型差异很大。

引用次数: 0

Integrating cross-view multi-scale perception and RAG-enabled expert fusion for medical prediction 集成跨视图多尺度感知和ragg支持的医学预测专家融合

IF 14.8

AI Open

Pub Date : 2026-01-01 Epub Date: 2026-02-27 DOI: 10.1016/j.aiopen.2026.02.004

Cheng Wang, Yongbin Liu, Ying Yu, Chunping Ouyang, Yaping Wan

Electronic Health Records (EHRs) continuously monitor patients’ health status in Intensive Care Units (ICUs), capturing irregular numerical time-series data and unstructured clinical text. While existing studies primarily focus on handling modality irregularities, they often overlook the complex intra- and inter-sequence interactions as well as the dependencies between short-term and long-term features. Moreover, clinical notes are typically semantically sparse and structurally noisy, making them difficult to interpret. To address these challenges, we propose a novel multimodal predictive model. For irregular numerical time-series data, we design a cross-view multi-scale framework that integrates cross-attention mechanisms with multi-scale convolutions. This enables dynamic modeling of diverse temporal embeddings while precisely capturing intrinsic inter-variable interactions and cross-temporal dependencies, all with reduced computational complexity. For clinical text, we adopt a retrieval-augmented technique that leverages external medical knowledge graphs (KGs) and large language models (LLMs) to enrich text representations related to medical codes. These enhanced embeddings are then fused with clinical notes via a gated mechanism, effectively alleviating semantic sparsity. We validate the effectiveness of the proposed approach on two critical clinical prediction tasks. Experimental results show maximum relative F1 score improvements of 3.3%, 6.0%, and 3.4% for MISTS, clinical notes, and multimodal fusion tasks, respectively, demonstrating our method’s excellent medical predictive capability.

电子健康记录（EHRs）持续监测重症监护病房（icu）患者的健康状况，捕获不规则的数值时间序列数据和非结构化临床文本。现有的研究主要集中在处理模态不规则性上，往往忽略了复杂的序列内和序列间的相互作用以及短期和长期特征之间的依赖关系。此外，临床记录通常语义稀疏，结构嘈杂，使其难以解释。为了解决这些挑战，我们提出了一种新的多模态预测模型。对于不规则数值时间序列数据，我们设计了一个跨视图多尺度框架，该框架将交叉注意机制与多尺度卷积相结合。这使得不同时间嵌入的动态建模成为可能，同时精确捕获内在的变量间相互作用和跨时间依赖性，所有这些都降低了计算复杂性。对于临床文本，我们采用了一种检索增强技术，该技术利用外部医学知识图（KGs）和大型语言模型（llm）来丰富与医学代码相关的文本表示。这些增强的嵌入然后通过门控机制与临床记录融合，有效缓解语义稀疏性。我们在两个关键的临床预测任务上验证了所提出方法的有效性。实验结果显示，对于mist、临床笔记和多模式融合任务，最大相对F1分数分别提高了3.3%、6.0%和3.4%，证明了我们的方法具有出色的医学预测能力。

{"title":"Integrating cross-view multi-scale perception and RAG-enabled expert fusion for medical prediction","authors":"Cheng Wang, Yongbin Liu, Ying Yu, Chunping Ouyang, Yaping Wan","doi":"10.1016/j.aiopen.2026.02.004","DOIUrl":"10.1016/j.aiopen.2026.02.004","url":null,"abstract":"<div><div>Electronic Health Records (EHRs) continuously monitor patients’ health status in Intensive Care Units (ICUs), capturing irregular numerical time-series data and unstructured clinical text. While existing studies primarily focus on handling modality irregularities, they often overlook the complex intra- and inter-sequence interactions as well as the dependencies between short-term and long-term features. Moreover, clinical notes are typically semantically sparse and structurally noisy, making them difficult to interpret. To address these challenges, we propose a novel multimodal predictive model. For irregular numerical time-series data, we design a cross-view multi-scale framework that integrates cross-attention mechanisms with multi-scale convolutions. This enables dynamic modeling of diverse temporal embeddings while precisely capturing intrinsic inter-variable interactions and cross-temporal dependencies, all with reduced computational complexity. For clinical text, we adopt a retrieval-augmented technique that leverages external medical knowledge graphs (KGs) and large language models (LLMs) to enrich text representations related to medical codes. These enhanced embeddings are then fused with clinical notes via a gated mechanism, effectively alleviating semantic sparsity. We validate the effectiveness of the proposed approach on two critical clinical prediction tasks. Experimental results show maximum relative F1 score improvements of 3.3%, 6.0%, and 3.4% for MISTS, clinical notes, and multimodal fusion tasks, respectively, demonstrating our method’s excellent medical predictive capability.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"7 ","pages":"Pages 62-70"},"PeriodicalIF":14.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147404238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Bio-inspired adaptive neurons for dynamic weighting in Artificial Neural Networks 人工神经网络中动态加权的仿生自适应神经元

IF 14.8

AI Open

Pub Date : 2026-01-01 Epub Date: 2026-02-05 DOI: 10.1016/j.aiopen.2026.02.001

Ashhadul Islam , Abdesselam Bouzerdoum , Samir Brahim Belhaouari

Traditional neural networks employ fixed weights during inference, limiting their ability to adapt to changing input conditions, unlike biological neurons that adjust signal strength dynamically based on stimuli. This discrepancy between artificial and biological neurons constrains neural network flexibility and adaptability. To bridge this gap, we propose a novel framework for adaptive neural networks, where neuron weights are modeled as functions of the input signal, allowing the network to adjust dynamically in real-time. Importantly, we achieve this within the same traditional architecture of an Artificial Neural Network, maintaining structural familiarity while introducing dynamic adaptability. In our research, we apply Chebyshev polynomials as one of the many possible decomposition methods to achieve this adaptive weighting mechanism, with polynomial coefficients learned during training. Of the 145 datasets tested, our adaptive Chebyshev neural network demonstrated a marked improvement over an equivalent MLP in approximately 83% of the cases, performing strictly better on 121 datasets. In the remaining 24 datasets, the performance of our algorithm matched that of the MLP, highlighting its ability to generalize the behavior of standard neural networks while offering enhanced adaptability. As a generalized form of MLP, this model seamlessly retains MLP performance where needed while extending its capabilities to achieve superior accuracy across a wide range of complex tasks. These results underscore the potential of adaptive neurons to enhance generalization, flexibility, and robustness in neural networks, particularly in applications with dynamic or non-linear data dependencies.

传统的神经网络在推理过程中使用固定的权重，限制了它们适应不断变化的输入条件的能力，不像生物神经元根据刺激动态调整信号强度。人工神经元与生物神经元的这种差异制约了神经网络的灵活性和适应性。为了弥补这一差距，我们提出了一种新的自适应神经网络框架，其中神经元权重被建模为输入信号的函数，允许网络实时动态调整。重要的是，我们在人工神经网络的相同传统架构中实现了这一点，在引入动态适应性的同时保持了结构熟悉度。在我们的研究中，我们使用切比雪夫多项式作为许多可能的分解方法之一来实现这种自适应加权机制，多项式系数在训练过程中学习。在测试的145个数据集中，我们的自适应Chebyshev神经网络在大约83%的情况下比等效MLP有显着改善，在121个数据集上表现得更好。在剩下的24个数据集中，我们的算法的性能与MLP相匹配，突出了其概括标准神经网络行为的能力，同时提供了增强的适应性。作为MLP的一种广义形式，该模型在需要的地方无缝地保留了MLP的性能，同时扩展了其能力，以在广泛的复杂任务中实现卓越的准确性。这些结果强调了自适应神经元在增强神经网络的泛化、灵活性和鲁棒性方面的潜力，特别是在动态或非线性数据依赖的应用中。

{"title":"Bio-inspired adaptive neurons for dynamic weighting in Artificial Neural Networks","authors":"Ashhadul Islam , Abdesselam Bouzerdoum , Samir Brahim Belhaouari","doi":"10.1016/j.aiopen.2026.02.001","DOIUrl":"10.1016/j.aiopen.2026.02.001","url":null,"abstract":"<div><div>Traditional neural networks employ fixed weights during inference, limiting their ability to adapt to changing input conditions, unlike biological neurons that adjust signal strength dynamically based on stimuli. This discrepancy between artificial and biological neurons constrains neural network flexibility and adaptability. To bridge this gap, we propose a novel framework for adaptive neural networks, where neuron weights are modeled as functions of the input signal, allowing the network to adjust dynamically in real-time. Importantly, we achieve this within the same traditional architecture of an Artificial Neural Network, maintaining structural familiarity while introducing dynamic adaptability. In our research, we apply Chebyshev polynomials as one of the many possible decomposition methods to achieve this adaptive weighting mechanism, with polynomial coefficients learned during training. Of the 145 datasets tested, our adaptive Chebyshev neural network demonstrated a marked improvement over an equivalent MLP in approximately 83% of the cases, performing strictly better on 121 datasets. In the remaining 24 datasets, the performance of our algorithm matched that of the MLP, highlighting its ability to generalize the behavior of standard neural networks while offering enhanced adaptability. As a generalized form of MLP, this model seamlessly retains MLP performance where needed while extending its capabilities to achieve superior accuracy across a wide range of complex tasks. These results underscore the potential of adaptive neurons to enhance generalization, flexibility, and robustness in neural networks, particularly in applications with dynamic or non-linear data dependencies.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"7 ","pages":"Pages 1-17"},"PeriodicalIF":14.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147404235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Not another imputation method: A transformer-based model for missing values in tabular datasets 不是另一种输入方法：基于转换器的模型，用于表格数据集中的缺失值

IF 14.8

AI Open

Pub Date : 2026-01-01 Epub Date: 2026-03-06 DOI: 10.1016/j.aiopen.2026.02.005

Camillo Maria Caruso , Paolo Soda , Valerio Guarrasi

Handling missing values in tabular datasets presents a significant challenge in training and testing artificial intelligence models, an issue usually addressed using imputation techniques. Here we introduce “Not Another Imputation Method” (NAIM), a novel transformer-based model specifically designed to address this issue without the need for traditional imputation techniques. NAIM’s ability to avoid the necessity of imputing missing values and to effectively learn from available data relies on two main techniques: the use of feature-specific embeddings to encode both categorical and numerical features also handling missing inputs; the modification of the masked self-attention mechanism to completely mask out the contributions of missing data. Additionally, a novel regularization technique is introduced to enhance the model’s generalization capability from incomplete data. We extensively evaluated NAIM on 5 publicly available tabular datasets, demonstrating its superior performance over 6 state-of-the-art machine learning models and 5 deep learning models, each paired with 3 different imputation techniques when necessary. The results highlight the efficacy of NAIM in improving predictive performance and resilience in the presence of missing data. To facilitate further research and practical application in handling missing data without traditional imputation methods, we made the code for NAIM available at https://github.com/cosbidev/NAIM.

处理表格数据集中的缺失值在训练和测试人工智能模型中提出了重大挑战，这个问题通常使用imputation技术来解决。在这里，我们介绍了“非另一种输入方法”（NAIM），这是一种新的基于变压器的模型，专门用于解决这一问题，而不需要传统的输入技术。NAIM避免输入缺失值的必要性并有效地从可用数据中学习的能力依赖于两种主要技术：使用特定于特征的嵌入来编码分类和数值特征，同时处理缺失的输入；对掩蔽自注意机制的修改，完全掩盖了缺失数据的贡献。此外，引入了一种新的正则化技术来提高模型对不完全数据的泛化能力。我们在5个公开可用的表格数据集上广泛评估了NAIM，证明了它在6个最先进的机器学习模型和5个深度学习模型上的卓越性能，每个模型在必要时都与3种不同的imputation技术配对。结果突出了NAIM在存在缺失数据的情况下提高预测性能和弹性的功效。为了便于进一步的研究和实际应用，在没有传统的补全方法的情况下处理缺失数据，我们在https://github.com/cosbidev/NAIM上提供了NAIM的代码。

{"title":"Not another imputation method: A transformer-based model for missing values in tabular datasets","authors":"Camillo Maria Caruso , Paolo Soda , Valerio Guarrasi","doi":"10.1016/j.aiopen.2026.02.005","DOIUrl":"10.1016/j.aiopen.2026.02.005","url":null,"abstract":"<div><div>Handling missing values in tabular datasets presents a significant challenge in training and testing artificial intelligence models, an issue usually addressed using imputation techniques. Here we introduce “Not Another Imputation Method” (NAIM), a novel transformer-based model specifically designed to address this issue without the need for traditional imputation techniques. NAIM’s ability to avoid the necessity of imputing missing values and to effectively learn from available data relies on two main techniques: the use of feature-specific embeddings to encode both categorical and numerical features also handling missing inputs; the modification of the masked self-attention mechanism to completely mask out the contributions of missing data. Additionally, a novel regularization technique is introduced to enhance the model’s generalization capability from incomplete data. We extensively evaluated NAIM on 5 publicly available tabular datasets, demonstrating its superior performance over 6 state-of-the-art machine learning models and 5 deep learning models, each paired with 3 different imputation techniques when necessary. The results highlight the efficacy of NAIM in improving predictive performance and resilience in the presence of missing data. To facilitate further research and practical application in handling missing data without traditional imputation methods, we made the code for NAIM available at <span><span>https://github.com/cosbidev/NAIM</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"7 ","pages":"Pages 96-122"},"PeriodicalIF":14.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147449778","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Knowledge intensive agents 知识密集型代理

IF 14.8

AI Open

Pub Date : 2026-01-01 Epub Date: 2026-02-05 DOI: 10.1016/j.aiopen.2026.02.002

Zhenghao Liu , Pengcheng Huang , Zhipeng Xu , Xinze Li , Shuliang Liu , Chunyi Peng , Haidong Xin , Yukun Yan , Shuo Wang , Xu Han , Zhiyuan Liu , Maosong Sun , Yu Gu , Ge Yu

Large Language Models (LLMs) have exhibited impressive capabilities in reasoning and language understanding. However, their reliance on memorized knowledge and tendency to generate hallucinated content limit their reliability in real-world applications. Retrieval-Augmented Generation (RAG) mitigates these issues by integrating a retrieval module that supplements LLMs with relevant external knowledge. This paradigm bridges parametric memory and explicit retrieval, offering a principled way to ground generation in factual evidence. Despite substantial progress, most prior work has focused on optimizing isolated components, either retrieval or generation, while overlooking the agentic perspective, in which LLMs act as autonomous agents capable of actively acquiring and strategically utilizing knowledge. In this perspectives paper, we argue for reinterpreting RAG as a collaborative knowledge process among agents with distinct yet complementary roles. We categorize knowledge-intensive agents into two primary roles: knowledge acquisition (e.g., routing, query reformulation) and knowledge utilization (e.g., knowledge refinement, response generation). From this viewpoint, RAG becomes a dynamic system in which knowledge is continuously transmitted, transformed, and aligned across agent roles. To fully realize this paradigm, we advocate a joint optimization framework for knowledge-intensive agents within RAG systems. This framework explicitly models the dynamics of knowledge flow in multi-agent settings, aligning knowledge supply with knowledge demand through LLM-driven data synthesis, feedback, and evaluation. By fostering adaptive and targeted knowledge exchange, the framework mitigates conflicts between parametric and retrieved knowledge, thereby enhancing both coherence and factuality. We argue that this multi-agent joint optimization paradigm improves RAG systems in scalability, reliability, and adaptability, unlocking the potential for next-generation knowledge-intensive LLMs that reason, retrieve, and collaborate across deep retrieval processes and diverse vertical domains.

大型语言模型（llm）在推理和语言理解方面表现出了令人印象深刻的能力。然而，它们对记忆知识的依赖和产生幻觉内容的倾向限制了它们在实际应用中的可靠性。检索增强生成（RAG）通过集成检索模块来缓解这些问题，该模块为llm补充了相关的外部知识。这种范式连接了参数记忆和显式检索，为事实证据的生成提供了一种原则性的方法。尽管取得了实质性的进展，但大多数先前的工作都集中在优化孤立的组件，无论是检索还是生成，而忽略了代理的角度，其中llm作为能够主动获取和战略性地利用知识的自主代理。在这篇观点论文中，我们主张将RAG重新解释为具有不同但互补角色的代理之间的协作知识过程。我们将知识密集型代理分为两个主要角色：知识获取（例如，路由，查询重新表述）和知识利用（例如，知识提炼，响应生成）。从这个角度来看，RAG成为一个动态系统，在这个系统中，知识在各个代理角色之间不断地传递、转换和对齐。为了充分实现这一范式，我们提倡为RAG系统中的知识密集型代理建立一个联合优化框架。该框架明确地为多智能体设置中的知识流动动态建模，通过llm驱动的数据合成、反馈和评估，使知识供给与知识需求保持一致。通过促进适应性和针对性的知识交换，该框架减轻了参数知识和检索知识之间的冲突，从而增强了一致性和真实性。我们认为，这种多智能体联合优化范例提高了RAG系统的可扩展性、可靠性和适应性，释放了下一代知识密集型llm的潜力，这些llm可以跨深度检索过程和不同的垂直领域进行推理、检索和协作。

{"title":"Knowledge intensive agents","authors":"Zhenghao Liu , Pengcheng Huang , Zhipeng Xu , Xinze Li , Shuliang Liu , Chunyi Peng , Haidong Xin , Yukun Yan , Shuo Wang , Xu Han , Zhiyuan Liu , Maosong Sun , Yu Gu , Ge Yu","doi":"10.1016/j.aiopen.2026.02.002","DOIUrl":"10.1016/j.aiopen.2026.02.002","url":null,"abstract":"<div><div>Large Language Models (LLMs) have exhibited impressive capabilities in reasoning and language understanding. However, their reliance on memorized knowledge and tendency to generate hallucinated content limit their reliability in real-world applications. Retrieval-Augmented Generation (RAG) mitigates these issues by integrating a retrieval module that supplements LLMs with relevant external knowledge. This paradigm bridges parametric memory and explicit retrieval, offering a principled way to ground generation in factual evidence. Despite substantial progress, most prior work has focused on optimizing isolated components, either retrieval or generation, while overlooking the agentic perspective, in which LLMs act as autonomous agents capable of actively acquiring and strategically utilizing knowledge. In this perspectives paper, we argue for reinterpreting RAG as a collaborative knowledge process among agents with distinct yet complementary roles. We categorize knowledge-intensive agents into two primary roles: knowledge acquisition (e.g., routing, query reformulation) and knowledge utilization (e.g., knowledge refinement, response generation). From this viewpoint, RAG becomes a dynamic system in which knowledge is continuously transmitted, transformed, and aligned across agent roles. To fully realize this paradigm, we advocate a joint optimization framework for knowledge-intensive agents within RAG systems. This framework explicitly models the dynamics of knowledge flow in multi-agent settings, aligning knowledge supply with knowledge demand through LLM-driven data synthesis, feedback, and evaluation. By fostering adaptive and targeted knowledge exchange, the framework mitigates conflicts between parametric and retrieved knowledge, thereby enhancing both coherence and factuality. We argue that this multi-agent joint optimization paradigm improves RAG systems in scalability, reliability, and adaptability, unlocking the potential for next-generation knowledge-intensive LLMs that reason, retrieve, and collaborate across deep retrieval processes and diverse vertical domains.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"7 ","pages":"Pages 18-44"},"PeriodicalIF":14.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147404236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SafeCast: Risk-responsive motion forecasting for autonomous vehicles SafeCast：自动驾驶汽车的风险响应运动预测

IF 14.8

AI Open

Pub Date : 2025-01-01 Epub Date: 2025-08-27 DOI: 10.1016/j.aiopen.2025.08.001

Haicheng Liao , Hanlin Kong , Zhenning Li , Chengzhong Xu

Accurate motion forecasting is essential for the safety and reliability of autonomous driving (AD) systems. While existing methods have made significant progress, they often overlook explicit safety constraints and struggle to capture the complex interactions among traffic agents, environmental factors, and motion dynamics. To address these challenges, we present SafeCast, a risk-responsive motion forecasting model that integrates safety-aware decision-making with uncertainty-aware adaptability. SafeCast is the first to incorporate the Responsibility-Sensitive Safety (RSS) framework into motion forecasting, encoding interpretable safety rules — such as safe distances and collision avoidance — based on traffic norms and physical principles. To further enhance robustness, we introduce the Graph Uncertainty Feature (GUF), a graph-based module that injects learnable noise into Graph Attention Networks, capturing real-world uncertainties and enhancing generalization across diverse scenarios. We evaluate SafeCast on four real-world benchmark datasets — Next Generation Simulation (NGSIM), Highway Drone (HighD), ApolloScape, and the Macao Connected Autonomous Driving (MoCAD) — covering highway, urban, and mixed-autonomy traffic environments. Our model achieves state-of-the-art (SOTA) accuracy while maintaining a lightweight architecture and low inference latency, underscoring its potential for real-time deployment in safety-critical AD systems.

准确的运动预测对于自动驾驶（AD）系统的安全性和可靠性至关重要。虽然现有的方法已经取得了重大进展，但它们往往忽略了明确的安全约束，并且难以捕捉交通代理、环境因素和运动动力学之间复杂的相互作用。为了应对这些挑战，我们提出了SafeCast，这是一种风险响应运动预测模型，将安全意识决策与不确定性意识适应性相结合。SafeCast是第一个将责任敏感安全（RSS）框架纳入运动预测的系统，根据交通规范和物理原理对可解释的安全规则（如安全距离和避免碰撞）进行编码。为了进一步增强鲁棒性，我们引入了图形不确定性特征（GUF），这是一个基于图形的模块，它将可学习噪声注入到图形注意网络中，捕捉现实世界的不确定性并增强不同场景的泛化。我们在四个真实世界的基准数据集——下一代模拟（NGSIM）、高速公路无人机（HighD）、ApolloScape和澳门互联自动驾驶（MoCAD）上对SafeCast进行了评估，涵盖高速公路、城市和混合自动驾驶交通环境。我们的模型实现了最先进的（SOTA）精度，同时保持了轻量级架构和低推理延迟，强调了其在安全关键型AD系统中实时部署的潜力。

{"title":"SafeCast: Risk-responsive motion forecasting for autonomous vehicles","authors":"Haicheng Liao , Hanlin Kong , Zhenning Li , Chengzhong Xu","doi":"10.1016/j.aiopen.2025.08.001","DOIUrl":"10.1016/j.aiopen.2025.08.001","url":null,"abstract":"<div><div>Accurate motion forecasting is essential for the safety and reliability of autonomous driving (AD) systems. While existing methods have made significant progress, they often overlook explicit safety constraints and struggle to capture the complex interactions among traffic agents, environmental factors, and motion dynamics. To address these challenges, we present SafeCast, a risk-responsive motion forecasting model that integrates safety-aware decision-making with uncertainty-aware adaptability. SafeCast is the first to incorporate the Responsibility-Sensitive Safety (RSS) framework into motion forecasting, encoding interpretable safety rules — such as safe distances and collision avoidance — based on traffic norms and physical principles. To further enhance robustness, we introduce the Graph Uncertainty Feature (GUF), a graph-based module that injects learnable noise into Graph Attention Networks, capturing real-world uncertainties and enhancing generalization across diverse scenarios. We evaluate SafeCast on four real-world benchmark datasets — Next Generation Simulation (NGSIM), Highway Drone (HighD), ApolloScape, and the Macao Connected Autonomous Driving (MoCAD) — covering highway, urban, and mixed-autonomy traffic environments. Our model achieves state-of-the-art (SOTA) accuracy while maintaining a lightweight architecture and low inference latency, underscoring its potential for real-time deployment in safety-critical AD systems.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"6 ","pages":"Pages 118-129"},"PeriodicalIF":14.8,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144907874","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

AI-generated content in landscape architecture: A survey 景观建筑中的ai生成内容：调查

IF 14.8

AI Open

Pub Date : 2025-01-01 Epub Date: 2025-10-22 DOI: 10.1016/j.aiopen.2025.10.002

Yue Xing , Wensheng Gan , Qidi Chen , Philip S. Yu

Landscape design is a complex process that requires designers to engage in intricate planning, analysis, and decision-making. This process involves the integration and reconstruction of science, art, and technology. Traditional landscape design methods are shaped by various factors, including the designer’s knowledge, time constraints, local ecological climate, available resources, and environmental considerations. These methods often rely on the designer’s personal experience and subjective aesthetics, with design standards rooted in subjective perception. As a result, they lack scientific and objective evaluation criteria and systematic design processes. Data-driven artificial intelligence (AI) technology provides an objective and rational design process. With the rapid development of different AI technologies, AI-generated content (AIGC) has permeated various aspects of landscape design at an unprecedented speed, serving as an innovative design tool. This article aims to explore the applications and opportunities of AIGC in landscape design. AIGC can support landscape design in areas such as site research and analysis, design concepts and scheme generation, parametric design optimization, plant selection and visual simulation, construction management, and process optimization. However, AIGC also faces challenges in landscape design, including data quality and reliability, design expertise and judgment, technical challenges and limitations, site characteristics and sustainability, user needs and participation, the balance between technology and creativity, ethics, and social impact. Finally, this article provides a detailed outlook on the future development trends and prospects of AIGC in landscape design. Through in-depth research and exploration in this review, readers can gain a better understanding of the relevant applications, potential opportunities, and key challenges of AIGC in landscape design.

景观设计是一个复杂的过程，需要设计师参与复杂的规划、分析和决策。这个过程涉及到科学、艺术和技术的整合和重建。传统的景观设计方法受到多种因素的影响，包括设计师的知识、时间限制、当地生态气候、可用资源和环境考虑。这些方法往往依赖于设计师的个人经验和主观审美，设计标准根植于主观感知。因此，缺乏科学客观的评价标准和系统的设计过程。数据驱动的人工智能（AI）技术提供了一个客观、理性的设计过程。随着各种人工智能技术的快速发展，人工智能生成内容（AI -generated content， AIGC）以前所未有的速度渗透到景观设计的各个方面，成为一种创新的设计工具。本文旨在探讨AIGC在景观设计中的应用和机遇。AIGC可以在场地研究和分析、设计概念和方案生成、参数化设计优化、植物选择和视觉模拟、施工管理和过程优化等领域支持景观设计。然而，AIGC在景观设计方面也面临着挑战，包括数据质量和可靠性、设计专业知识和判断力、技术挑战和局限性、场地特征和可持续性、用户需求和参与、技术与创意、道德和社会影响之间的平衡。最后，对AIGC在景观设计中的未来发展趋势和前景进行了详细的展望。通过本文的深入研究和探索，读者可以更好地了解AIGC在景观设计中的相关应用、潜在机遇和主要挑战。

{"title":"AI-generated content in landscape architecture: A survey","authors":"Yue Xing , Wensheng Gan , Qidi Chen , Philip S. Yu","doi":"10.1016/j.aiopen.2025.10.002","DOIUrl":"10.1016/j.aiopen.2025.10.002","url":null,"abstract":"<div><div>Landscape design is a complex process that requires designers to engage in intricate planning, analysis, and decision-making. This process involves the integration and reconstruction of science, art, and technology. Traditional landscape design methods are shaped by various factors, including the designer’s knowledge, time constraints, local ecological climate, available resources, and environmental considerations. These methods often rely on the designer’s personal experience and subjective aesthetics, with design standards rooted in subjective perception. As a result, they lack scientific and objective evaluation criteria and systematic design processes. Data-driven artificial intelligence (AI) technology provides an objective and rational design process. With the rapid development of different AI technologies, AI-generated content (AIGC) has permeated various aspects of landscape design at an unprecedented speed, serving as an innovative design tool. This article aims to explore the applications and opportunities of AIGC in landscape design. AIGC can support landscape design in areas such as site research and analysis, design concepts and scheme generation, parametric design optimization, plant selection and visual simulation, construction management, and process optimization. However, AIGC also faces challenges in landscape design, including data quality and reliability, design expertise and judgment, technical challenges and limitations, site characteristics and sustainability, user needs and participation, the balance between technology and creativity, ethics, and social impact. Finally, this article provides a detailed outlook on the future development trends and prospects of AIGC in landscape design. Through in-depth research and exploration in this review, readers can gain a better understanding of the relevant applications, potential opportunities, and key challenges of AIGC in landscape design.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"6 ","pages":"Pages 220-243"},"PeriodicalIF":14.8,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145415041","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

FedGPA: Federated Learning with Global Personalized Aggregation FedGPA：具有全球个性化聚合功能的联合学习

AI Open

Pub Date : 2025-01-01 Epub Date: 2025-04-15 DOI: 10.1016/j.aiopen.2025.03.001

Zongfu Han , Yu Feng , Yifan Zhu , Zhen Tian , Fangyu Hao , Meina Song

A significant challenge in Federated Learning (FL) is addressing the heterogeneity of local data distribution across clients. Personalized Federated Learning (PFL), an emerging method aimed at overcoming data heterogeneity, can either integrate personalized components into the global model or train multiple models to achieve personalization. However, little research has simultaneously considered both directions. One such approach involves adopting a weighted aggregation method to generate personalized models, where the weights are determined by solving an optimization problem among different clients. In brief, previous works either neglect the use of global information during local representation learning or simply treat the personalized model as learning a set of individual weights. In this work, we initially decouple the model into a feature extractor, associated with generalization, and a classifier, linked to personalization. Subsequently, we conduct local–global alignment based on prototypes to leverage global information for learning better representations. Moreover, we fully utilize these representations to calculate the distance between clients and develop individual aggregation strategies for feature extractors and classifiers, respectively. Finally, extensive experimental results on five benchmark datasets under three different heterogeneous data scenarios demonstrate the effectiveness of our proposed FedGPA.

联邦学习（FL）的一个重大挑战是解决客户端本地数据分布的异构性。个性化联邦学习（PFL）是一种旨在克服数据异质性的新兴方法，它可以将个性化组件集成到全局模型中，也可以训练多个模型来实现个性化。然而，很少有研究同时考虑这两个方向。其中一种方法涉及采用加权聚合方法来生成个性化模型，其中权重是通过解决不同客户端的优化问题来确定的。简而言之，以前的工作要么忽略了局部表示学习过程中全局信息的使用，要么简单地将个性化模型视为学习一组个体权重。在这项工作中，我们最初将模型解耦为与泛化相关的特征提取器和与个性化相关的分类器。随后，我们进行基于原型的局部-全局对齐，以利用全局信息来学习更好的表示。此外，我们充分利用这些表示来计算客户端之间的距离，并分别为特征提取器和分类器开发单独的聚合策略。最后，在3种不同异构数据场景下的5个基准数据集上进行了大量的实验，验证了我们所提出的FedGPA算法的有效性。

{"title":"FedGPA: Federated Learning with Global Personalized Aggregation","authors":"Zongfu Han , Yu Feng , Yifan Zhu , Zhen Tian , Fangyu Hao , Meina Song","doi":"10.1016/j.aiopen.2025.03.001","DOIUrl":"10.1016/j.aiopen.2025.03.001","url":null,"abstract":"<div><div>A significant challenge in Federated Learning (FL) is addressing the heterogeneity of local data distribution across clients. Personalized Federated Learning (PFL), an emerging method aimed at overcoming data heterogeneity, can either integrate personalized components into the global model or train multiple models to achieve personalization. However, little research has simultaneously considered both directions. One such approach involves adopting a weighted aggregation method to generate personalized models, where the weights are determined by solving an optimization problem among different clients. In brief, previous works either neglect the use of global information during local representation learning or simply treat the personalized model as learning a set of individual weights. In this work, we initially decouple the model into a feature extractor, associated with generalization, and a classifier, linked to personalization. Subsequently, we conduct local–global alignment based on prototypes to leverage global information for learning better representations. Moreover, we fully utilize these representations to calculate the distance between clients and develop individual aggregation strategies for feature extractors and classifiers, respectively. Finally, extensive experimental results on five benchmark datasets under three different heterogeneous data scenarios demonstrate the effectiveness of our proposed FedGPA.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"6 ","pages":"Pages 82-92"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143847949","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Proactive Recommendation in Social Networks: Steering user interest with causal inference 社交网络中的主动推荐：用因果推理引导用户兴趣

IF 14.8

AI Open

Pub Date : 2025-01-01 Epub Date: 2025-09-02 DOI: 10.1016/j.aiopen.2025.08.003

Hang Pan , Shuxian Bi , Wenjie Wang , Haoxuan Li , Peng Wu , Fuli Feng

Recommending items that solely cater to users’ historical interests narrows users’ horizons. Recent works have considered steering target users beyond their historical interests by directly adjusting items exposed to them. However, the recommended items for direct steering might not align perfectly with the evolution of users’ interests, detrimentally affecting the target users’ experience.

To avoid this issue, we propose a new task named Proactive Recommendation in Social Networks (PRSN) that indirectly steers users’ interest by utilizing the influence of social neighbors, i.e.,indirect steering by adjusting the exposure of a target item to target users’ neighbors. The key to PRSN lies in answering an interventional question: what would a target user’s feedback be on a target item if the item is exposed to the user’s different neighbors? To answer this question, we resort to causal inference and formalize PRSN as: (1) estimating the potential feedback of a user on an item, under the network interference by the item’s exposure to the user’s neighbors; and (2) adjusting the exposure of a target item to target users’ neighbors to trade-off steering performance and the damage to the neighbors’ experience. To this end, we propose a Neighbor Interference Recommendation (NIRec) framework with two modules: (1) an interference representation-based estimation module for modeling potential feedback; (2) a post-learning-based optimization module for adjusting a target item’s exposure to trade-off steering performance and the neighbors’ experience through greedy search. We conduct extensive semi-simulation experiments on real-world datasets, validating the steering effectiveness of NIRec. The code is available at https://github.com/HungPaan/NIRec.

只推荐迎合用户历史兴趣的商品会缩小用户的视野。最近的作品考虑通过直接调整暴露在目标用户面前的项目来引导目标用户超越他们的历史兴趣。然而，直接引导的推荐项目可能与用户兴趣的演变不完全一致，从而对目标用户的体验产生不利影响。为了避免这一问题，我们提出了一种新的任务，即主动推荐（PRSN），它通过利用社会邻居的影响间接引导用户的兴趣，即通过调整目标项目对目标用户邻居的曝光来间接引导。PRSN的关键在于回答一个干涉性问题：如果一个目标物品暴露给用户的不同邻居，目标用户对该物品的反馈会是什么？为了回答这个问题，我们采用因果推理，并将PRSN形式化为：(1)在网络干扰下，估计用户对物品的潜在反馈，该物品暴露于用户的邻居；(2)调整目标物品对目标用户邻居的暴露，以权衡转向性能和对邻居体验的损害。为此，我们提出了一个包含两个模块的邻居干扰推荐（NIRec）框架：(1)基于干扰表示的估计模块，用于建模潜在反馈；(2)基于后学习的优化模块，通过贪婪搜索调整目标项目对权衡转向性能的暴露和邻居的经验。我们在真实世界的数据集上进行了大量的半模拟实验，验证了NIRec的转向有效性。代码可在https://github.com/HungPaan/NIRec上获得。

{"title":"Proactive Recommendation in Social Networks: Steering user interest with causal inference","authors":"Hang Pan , Shuxian Bi , Wenjie Wang , Haoxuan Li , Peng Wu , Fuli Feng","doi":"10.1016/j.aiopen.2025.08.003","DOIUrl":"10.1016/j.aiopen.2025.08.003","url":null,"abstract":"<div><div>Recommending items that solely cater to users’ historical interests narrows users’ horizons. Recent works have considered steering target users beyond their historical interests by directly adjusting items exposed to them. However, the recommended items for direct steering might not align perfectly with the evolution of users’ interests, detrimentally affecting the target users’ experience.</div><div>To avoid this issue, we propose a new task named <em>Proactive Recommendation in Social Networks</em> (<strong>PRSN</strong>) that indirectly steers users’ interest by utilizing the influence of social neighbors, <em>i.e.,</em>indirect steering by adjusting the exposure of a target item to target users’ neighbors. The key to PRSN lies in answering an interventional question: <em>what would a target user’s feedback be on a target item if the item is exposed to the user’s different neighbors?</em> To answer this question, we resort to causal inference and formalize PRSN as: (1) estimating the potential feedback of a user on an item, under the network interference by the item’s exposure to the user’s neighbors; and (2) adjusting the exposure of a target item to target users’ neighbors to trade-off steering performance and the damage to the neighbors’ experience. To this end, we propose a <strong>N</strong>eighbor <strong>I</strong>nterference <strong>Rec</strong>ommendation (<strong>NIRec</strong>) framework with two modules: (1) an interference representation-based estimation module for modeling potential feedback; (2) a post-learning-based optimization module for adjusting a target item’s exposure to trade-off steering performance and the neighbors’ experience through greedy search. We conduct extensive semi-simulation experiments on real-world datasets, validating the steering effectiveness of NIRec. The code is available at <span><span>https://github.com/HungPaan/NIRec</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":100068,"journal":{"name":"AI Open","volume":"6 ","pages":"Pages 130-141"},"PeriodicalIF":14.8,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145120788","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0