Accepted workshops
Abstract
Modern networked systems generate massive volumes of heterogeneous and continuously evolving data, creating major challenges for data management, intelligent analysis, and cybersecurity. Traditional machine learning approaches often struggle to adapt to the scale, decentralization, and dynamic nature of contemporary network environments, motivating the need for adaptive, distributed, and knowledge-driven AI solutions. This workshop explores recent advances in AI-driven network intelligence and cybersecurity, with a particular focus on information and knowledge management techniques for distributed systems. Topics include Generative AI (GenAI) for synthetic network data generation and augmentation, Large Language Models (LLMs) and Natural Language Processing (NLP) for cybersecurity intelligence and threat analysis, continual and federated learning for privacy-preserving adaptive systems, adversarial learning and trustworthy AI, and agentic AI for autonomous and human-in-the-loop security workflows. The workshop also investigates retrieval-augmented generation (RAG), knowledge graphs, reasoning over structured and unstructured security data, intelligent orchestration, and scalable AI-enabled network analytics. By bringing together researchers and practitioners from AI, cybersecurity, networking, and knowledge management, the workshop aims to foster interdisciplinary discussion on resilient, intelligent, and trustworthy networked systems.
Abstract
This proposal outlines the organization of the 3rd Workshop on Evaluating and Applying Recommender Systems with Large Language Models (EARL), to be co-located with the 35th International ACM Conference on Knowledge and Information Management (CIKM 2026) in Rome, Italy. Following the strong engagement of the first two EARL editions, this third installment aims to solidify the workshop as a premier venue for advancing research at the intersection of large language models (LLMs) and recommender systems (RSs). EARL 2026 will foster discussions on the evaluation and practical deployment of LLM-driven RSs, emphasizing not only methodological soundness and reproducibility, but also the responsible integration of LLMs into real-world systems. By bringing together researchers and practitioners from academia and industry, the workshop will stimulate debate, encourage critical reflection, and shape future research directions in LLM-enhanced RSs.
Abstract
The increasing adoption of personalization and recommender systems in high-impact domains raises fundamental questions about how user models represent, reason about, and adapt to human behavior. While recent advances in machine learning have improved predictive accuracy, they offer limited support for explicit reasoning, interpretability, and the incorporation of cognitive and normative constraints. The Second International Workshop on Hybrid AI for Human-Centric Personalization and Recommendation (HyPeR 2026) focuses on hybrid AI as a methodological framework for user modeling, recommendations, and personalization, emphasizing the integration of learning-based approaches with symbolic knowledge, reasoning mechanisms, and cognition-informed representations. HyPeR brings together researchers and practitioners to examine architectures, methods, and evaluation strategies for hybrid user models. The workshop aims to stimulate exchange and to shape future research directions at the intersection of user modeling, hybrid AI, and human-centered recommender systems.
Abstract
The GREEN-AI workshop addresses the growing need for environmentally sustainable Artificial Intelligence. As AI models continue to scale in size and complexity, their energy consumption, carbon footprint, and resource demands raise urgent questions about the long-term ecological and social consequences of deploying large-scale AI systems at an ever-increasing pace. Distinct from workshops that apply AI to environmental or climate-related problems, GREEN-AI focuses specifically on the sustainability of AI itself—spanning its computational, methodological, and infrastructural dimensions. It offers a dedicated forum for researchers, practitioners, and policymakers to explore energy-efficient algorithms, sustainable data management and infrastructure practices, and novel evaluation metrics that explicitly incorporate environmental impact alongside traditional performance measures. By bridging academia and industry, GREEN-AI aims to catalyze actionable change and foster innovation that reconciles rapid technical advancement with global sustainability objectives. The workshop encourages contributions from diverse perspectives, including hardware-aware optimization, green computing, and responsible model development. Structured as a focused half-day event, it will feature invited talks, peer-reviewed paper presentations, and interactive panel discussions designed to shape a more responsible, resource-conscious, and ecologically aware AI ecosystem.
Organised by: Nida Meddouri, Juba Agoun, Clément Iphar, Aurélie Leborgne, Loïc Salmon, Nazha Selmaoui and Assaad Zeghina
Abstract
Spatial, temporal, and spatio-temporal data are now central to many CIKM-relevant applications, including mobility analytics, earth observation, environmental monitoring, maritime intelligence, healthcare trajectories, sensor platforms, geotagged text, and distributed information systems. Yet research on these data remains fragmented across domains, while key challenges such as heterogeneous data integration, scalable management, uncertainty handling, privacy, robust evaluation, explainability, and reproducibility are shared across applications. We propose STMA4KD, a half-day workshop dedicated to the management and analysis of spatial and temporal data across the full data lifecycle, from acquisition and integration to modeling, evaluation, and dissemination. The workshop will bring together researchers and practitioners from data management, knowledge discovery, information retrieval, machine learning, and various application domains. STMA4KD will combine peer-reviewed papers, lightning talks, poster interactions, and an open-problems session to foster active engagement and identify shared research priorities. Building on the GAST workshop series and related thematic scientific events (https://gt-gast.irisa.fr/actions/), STMA4KD aims to establish a broader international venue within CIKM for discussing methods, benchmarks, datasets, and future directions for spatial and temporal data.
Organised by: Gabriele Tolomei, Ziheng Chen, Flavio Giorgi, Vittoria Vineis, Matteo Silvestri, Fabiano Veglianti, Edoardo Gabrielli and Lorenzo Antonelli
Abstract
As AI systems are increasingly deployed in high-stakes domains, regulatory and societal demands for transparency and interpretability continue to grow. While classical Explainable AI (XAI) methods produce structured technical artifacts (e.g., feature attributions and counterfactuals examples), these remain difficult to interpret for non-technical users and only partially meet usability and legal requirements. Large Language Models (LLMs) can transform these structured outputs into accessible natural-language XAI narratives, but they introduce new challenges regarding faithfulness, hallucination, and epistemic misalignment. LLM4XAI sits at the intersection of XAI, NLP, IR, and HCI, investigating how generative systems can act as reliable mediators. The workshop provides a multidisciplinary forum for both academia and industry, focusing on the grounding, robustness, and real-world deployment of XAI narratives to ensure AI explanations bridge the gap between technical rigor and end-user accessibility.
Organised by: Mauro Barni, Irene Amerini, Gian Luca Marcialis, Roberto Caldelli, Lorenzo Cirillo, Claudio Schiavella, Simone Teglia, Pietro Bongini and Andrea Ciamarra
Abstract
The rapid growth of generative multimodal processing and dissemination technologies has enabled the creation of an unprecedented amount of realistic synthetic media [1]. As a result, methods for assessing the integrity and authenticity of digital media play a crucial role across scientific, industrial, and societal contexts, as the realism of altered digital content continues to challenge forensic models [5]. In this context, multimedia forensic systems must address factors such as social media preprocessing, heterogeneous data sources, distribution shifts, adversarial perturbations, and the trustworthiness of information, as well as disinformation detection. These conditions underscore the challenges of deploying digital content integrity tools that remain effective, robust, and explainable [4]. Therefore, this workshop will highlight research advancements in multimedia forensics, deepfake detection, adversarial machine learning, and AI-generated media attribution. Particular focus will be on the deployment of autonomous AI agents for media verification, as well as lifelong media authentication, a forward-looking perspective on AI’s role in safeguarding media integrity [2]. The workshop gathers interdisciplinary contributions that advance the foundations and practice of robust and trustworthy multimedia forensics across different modalities [3], such as images, text, and audio, encouraging research that challenges existing assumptions and proposes forward-looking solutions across multimedia analysis.
Organised by: Nora Nahr, Benedikt Elser, Michael Müller and Florian Wahl
Abstract
Much of the knowledge organizations rely on is implicit: it lives in the routines, judgement, and experience of individual experts, and it is rarely written down. As workforces turn over and experts retire, this knowledge is lost before it ever reaches a database or document collection. The information and knowledge management community has strong methods for extracting and organizing knowledge that already exists in text and data, but comparatively little attention is paid to the step before that, surfacing implicit knowledge from people and work practice so that it can be represented, integrated, and preserved at all. KEEPER focuses on this gap. The workshop brings together research on eliciting implicit and expert knowledge through interactive and AI-supported methods, externalizing it into structured and reusable representations, integrating it with existing knowledge management systems, and preserving it as durable organizational memory. We treat AI as assistive and human-in-the-loop throughout. KEEPER is a half-day hybrid workshop combining peer-reviewed contributions with an interactive session that produces a shared research agenda, targeting researchers and practitioners in knowledge management, information retrieval, NLP, and human-AI interaction.
Organised by: Ashish Jain, Shubham Chatterjee, Amit Goyal, Xi Wang, Marios Kokkodis, Panos Adamopoulos, Selen Uguroglu, Katerina Iliakopoulou-Zanos, Duong Tran and Minmin Chen
Abstract
Personalization is being rebuilt around generative models. Large language models, retrieval-augmented generation, and agentic systems now shape how recommenders retrieve, rank, explain, converse, and even simulate users—yet they also surface hard problems of hallucination, control, evaluation, privacy, and fairness. GRAIL 2026 brings together academic researchers and industry practitioners to advance generative, retrieval-augmented, and agentic approaches to recommendation—including LLM-based retrieval and ranking, retrieval-augmented and knowledge-grounded personalization, conversational and agentic recommenders, multimodal content generation, user simulation, and synthetic data—while giving equal attention to these open challenges and their long-term user impact.
Organised by: Antonio Galli, Michela Gravina, Antonino Ferraro, Martina Iammarino, Valerio Guarrasi, Angel Garcia-Pedrero, Consuelo Gonzalo-Martin, Paolo Soda, Vincenzo Moscato and Carlo Sansone
Abstract
Clinical decision-making increasingly depends on heterogeneous and distributed information sources, including medical images, radiology and pathology reports, electronic health records, laboratory measurements, biosignals, temporal patient trajectories, and clinical guidelines. Recent advances in multimodal learning, foundation models, and retrieval-augmented generation have opened new opportunities for Artificial Intelligence (AI) in clinical scenarios. However, current systems still struggle to support effective information access and knowledge integration across these sources, as retrieving, aligning, fusing, and reasoning over heterogeneous clinical evidence remains challenging in reliable, interpretable, and workflow-aware settings. The workshop Clinical-MIRF: Multimodal Information Retrieval and Fusion in Clinical Workflows aims to explore innovative methodologies and applications in multimodal information retrieval, cross-modal fusion, knowledge-grounded reasoning, and trustworthy information access in clinical workflows. Clinical-MIRF will bring together researchers from information retrieval, knowledge management, biomedical AI, medical imaging, natural language processing, clinical informatics, and human-centered AI. Topics include multimodal indexing and search, cross-modal representation learning, clinical Retrieval-Augmented Generation (RAG), knowledge graph-enhanced retrieval, fusion of imaging and textual evidence, evaluation protocols for multimodal clinical systems, robustness, fairness, privacy, and humam-in-the-loop clinical decision support. Accepted workshop papers will be published in post-proceedings by Springer in the Lecture Notes in Computer Science (LNCS) series.
Organised by: Anett Hoppe, Nilavra Bhattacharya, Jiqun Liu and Ran Yu
Abstract
The emergence of conversational AI, retrieval-augmented generation and adptive recommendation systems has fundamentally reshaped how learners access, evaluate and construct knowledge from the web. Yet our understanding of \textit{how} these shifts affect learning processes, and how retrieval systems should be designed to support them still remains limited. The sixth edition of the IWILDS workshop series brings together researchers from information retrieval, knowledge management, learning analytics, user modeling and educational psychology to address this gap. The workshop focuses on three interconnected research questions: how learners formulate and reformulate queries and prompts across search, RAG and conversational paradigms; what evaluation frameworks can capture learning outcomes in AI-mediated information environments; and how to construct datasets that reflect contemporary learning-in-the-wild. IWILDS’26 combines a paper-based morning session (paper presentations and invited talk) with an interactive afternoon focused on structured discussions about research questions and a collaboration fair. In its sixth year, the workshop builds on a sustained community anchored in IR and learning sciences, returning to CIKM with a sharpened focus aligned with CIKM’s core interests in information retrieval, user modeling and knowledge management.
Organised by: Giacomo Frisoni, Luca Ragazzi, Gianluca Moro, Zaiqiao Meng, Jinyuan Fang and Yufang Hou
Abstract
Modern information retrieval (IR) is built on pretrained language models whose parametric knowledge has driven strong gains in semantic understanding. Yet, this knowledge is frozen at training time, opaque, and shallow on long-tail, domain-specific, and fast-changing information, yielding imprecise, non-factual, or poorly grounded results. Knowledge-enhanced IR addresses this by augmenting retrieval systems—and the language models that power them—with external knowledge: unstructured or structured, gold or synthetic, increasingly across modalities. Large language models are now reshaping how this knowledge is acquired and integrated—for example, autonomous agents that plan, reason, and decide what to search; automated knowledge curation and linking; and systems that interleave retrieval and generation, learning when to consult external evidence and when to answer from parametric knowledge. The Third Workshop on Knowledge-Enhanced Information Retrieval (KEIR @ CIKM 2026) gathers researchers and practitioners from IR, natural language processing, data mining, and knowledge management to advance the principled integration of external knowledge into IR. KEIR welcomes both methodological and applied contributions, with particular attention to knowledge-intensive, domain-sensitive fields such as science, medicine, law, and finance, where factuality, trust, and discovery are paramount.
Abstract
Social simulation plays an important role in domains where human studies face practical constraints, such as sociology, psychology, medicine, and education. Conventional social simulations often rely on rule-based, statistical, or mechanistic models with predefined assumptions, which can oversimplify the complexity, diversity, and context dependence of real-world human behavior.
Recent advances in Large Language Models (LLMs) offer a promising opportunity by enabling agents to interpret contextual information, perform flexible reasoning, and adapt their behaviors to evolving environments, opening new frontiers for modeling human behavior, social interactions, and collective dynamics. Despite these opportunities, LLM-agent-based social simulations raise significant challenges, including persona and domain-knowledge modeling, behavioral evaluation and analysis, safety, ethics, and accountability in error-sensitive domains, and scalable infrastructure for large-scale and long-horizon simulations.
To answer these questions, the second LLM Agents for Social Simulation workshop (LASS), themed “Fundamental Challenges and Real-World Applications”, aims to bring together researchers and practitioners from diverse disciplines and backgrounds to explore how LLM agents can be more faithfully modeled, rigorously evaluated, safely governed, and effectively applied to address real-world problems in vertical domains, with careful consideration of their unique requirements, challenges, and practical limitations, thereby promoting human well-being and broader societal benefit.
Organised by: Jamila Smith-Loud, Andrew Smart, Lijun Qian and Xishuang Dong
Abstract
Recent advances in Generative AI (GenAI) have enabled strong capabilities in language understanding, content generation, vision-language reasoning, and multimodal interaction. However, most existing benchmarks emphasize general-purpose reasoning and task performance, offering limited evaluation of whether AI systems can accurately understand, represent, retrieve, and generate culturally grounded knowledge across diverse communities and social contexts. This limitation restricts the reliable deployment of GenAI in high-impact domains such as education, healthcare, and public services. The REACH-AI 2026 Workshop brings together researchers, industry leaders, practitioners, and community stakeholders to advance community-informed approaches for evaluating culturally competent Generative AI. Motivated by the REACH GenAI Consortium, a collaboration between Google Research and twelve Historically Black Colleges and Universities (HBCUs), the workshop focuses on culturally grounded datasets, community-informed taxonomies, evaluation frameworks, benchmark development, and human-centered assessment protocols for measuring cultural competency in AI systems. It aims to foster interdisciplinary dialogue across the CIKM, information retrieval, knowledge management, multimodal AI, and responsible AI communities. Through invited talks, paper presentations, and panel discussions, participants will explore methods for assessing cultural relevance, contextual understanding, representation, and knowledge utilization in GenAI systems. Expected outcomes include a community roadmap, collaborative research opportunities, and recommendations for future datasets, benchmarks, and evaluation standards for culturally competent AI.
Organised by: Martino Ciaperoni, Marco Minici, Vincenzo Moscato and Marco Postiglione
Abstract
Generative AI has become a general-purpose technology for producing text, images, code, and explanations across knowledge-intensive tasks. This opens new opportunities for social good in domains where expertise, data, and resources are scarce: rare diseases, low-resource education, accessibility, crisis response, and information integrity. Yet large-scale deployment also carries risks that fall hardest on disadvantaged communities, including hallucination, bias, privacy leakage, and high environmental cost. GenAI4SG brings together researchers working on retrieval-augmented generation, agentic systems, multimodal models, and knowledge-grounded AI to study how these technologies can be responsibly designed, evaluated, and deployed for social good, in line with the UN Sustainable Development Goals. We welcome technical, empirical, resource, demo, and position papers on methods, systems, datasets, benchmarks, and applications.
Organised by: Benyou Wang, Amit Kumar Jaiswal, Ruchir Gupta, Jiale Han, Prayag Tiwari, Amit Agarwal and Shandar Ahmad
Abstract
Large language models (LLMs) have demonstrated remarkable capabilities in natural language processing and other adaptable domains, yet they often struggle with knowledge grounding and reasoning. Integrating graph structures, such as knowledge graphs and graph neural networks offer a promising avenue to enhance LLMs with rich relational information. Our workshop \emph{Graph-Augmented LLMs} explores the synergistic convergence of LLMs and graph-based methodologies and its applications with its transformative potential. This integration facilitates advancements in areas such as enhanced reasoning, improved contextual understanding, and robust generalization. We aim to foster discussions on leveraging Graph Neural Networks (GNNs) and graph representation learning to augment LLMs, as well as investigating the application of federated learning and unlearning techniques to address privacy and ethical concerns in this rapidly evolving field. This workshop aims to foster discussions on how graph structures can empower LLMs to achieve improved reasoning, knowledge integration, and data privacy in diverse applications, including but not limited to biomedical, environmental and social-economic systems, with highly structured knowledge. The proposed workshop is the second iteration of GaLM 2025 workshop\footnote{\url{https://iitbhu.ac.in/cf/jcsic/activities/workshops}}, which was successfully organized at ICDM 2025 in Washington DC, USA.