WMAI 2026: Workshop on World Models and Autonomous Intelligence for Data Mining
In conjunction with IEEE ICDM 2026
About the Workshop
The concept of Autonomous Machine Intelligence (AMI), as articulated by Yann LeCun in his influential 2022 position paper, proposes a comprehensive architectural blueprint for building intelligent agents capable of learning, reasoning, and planning in ways that resemble human and animal cognition.
The AMI architecture integrates several key components: a configurable predictive world model, behavior driven through intrinsic motivation, and hierarchical Joint Embedding Predictive Architectures (H-JEPA) trained via self-supervised learning.
However, a significant gap remains between the theoretical vision of AMI and its practical realization in real-world data mining and AI systems. This workshop aims to bridge that gap by focusing specifically on practical implementations, empirical validations, and scalable deployments of AMI-inspired techniques in data mining.
Call for Papers (Topics of Interest)
We solicit original research papers, position papers, and system demonstrations on the following topics (but not limited to):
World Models for Data Mining
- âś“Predictive world models for temporal and spatio-temporal data mining
- âś“Configurable world models that adapt to different data mining tasks
- âś“Learning environment dynamics from unstructured and heterogeneous data
- âś“World model-based planning for automated data analytics pipelines
Joint Embedding Predictive Architectures (JEPA)
- âś“Practical implementations of I-JEPA, V-JEPA, and H-JEPA for mining tasks
- âś“Joint embedding methods for multi-modal data fusion and mining
- âś“Energy-based models for structured prediction in graphs, text, and time series
- âś“Self-supervised pre-training strategies inspired by JEPA for data mining
Intrinsic Motivation & Autonomous Exploration
- âś“Curiosity-driven exploration for data acquisition and active learning
- âś“Intrinsic reward mechanisms for reinforcement learning-based data mining
- âś“Autonomous feature discovery and representation learning
- âś“Self-directed learning strategies for mining in evolving environments
Scalable and Deployable AMI Systems
- âś“System architectures for real-time AMI-based analytics
- âś“Federated and distributed implementations of AMI components
- âś“Industrial case studies: AMI in recommendation, fraud detection, healthcare, smart cities
- âś“Benchmarks and evaluation frameworks for AMI-inspired data mining
Safety, Interpretability, and Ethics
- âś“Interpretable world models and explainable autonomous decision-making
- âś“Safety guarantees and alignment in AMI-driven systems
- âś“Ethical considerations in deploying autonomous data mining agents
- âś“Robustness and fairness in AMI-based predictions
Submission Guidelines
Paper Format
All submissions must be written in English and formatted according to the IEEE ICDM workshop paper template (IEEE 2-column format).
Submission Types
Original research with experimental validation.
Visionary ideas, preliminary results, or work-in-progress.
Demonstrations of practical AMI implementations.
Review Process: Each submission will be reviewed by at least 3 program committee members in a single-blind review process.
Publication: Accepted papers will be published in the IEEE ICDM Workshop Proceedings by the IEEE Computer Society Press and indexed in IEEE Xplore Digital Library.
Acknowledgment
The Microsoft CMT service was used for managing the peer-reviewing process for this conference. This service was provided for free by Microsoft and they bore all expenses, including costs for Azure cloud services as well as for software development and support.
Important Dates
| Milestone | Date |
|---|---|
| Workshop paper submission | Aug 20, 2026 |
| Acceptance notification | Sep 18, 2026 |
| Camera-ready due | Oct 2, 2026 |
| Workshop date (08:30 – 12:00, Room London) | Nov 12, 2026 |
Workshop Program
Session Chair: A/Prof. Hien Dinh Nguyen, Ph.D. (University of Information Technology, VNU-HCM, Vietnam)
Each paper is allotted 20 minutes: 15 minutes of presentation and 5 minutes of questions. All times are local (China Standard Time, UTC+8).
| 08:30 - 08:40 | Opening and welcome Prof. Fabrice Guillet (Nantes Université), online, on behalf of the Workshop Chairs |
| 08:40 - 09:50 | Session 1 – World Models and Autonomous Workflows Session chair: A/Prof. Hien Dinh Nguyen |
| 08:40 - 09:00 | S02203PRoVE: Separating Self-Motion from World Change in Video World Models via Camera-Pose Positional Embeddings Tsai Shuo Kuo, Masanori Yokota, Ryutaro Ichise |
| 09:00 - 09:20 | S02206Per-Step Proofs Do Not Prove a Roll-out: Temporal Integrity as a Missing Obligation for World Models Nhan Vo, Duy Ninh, Hiep Xuan Huynh |
| 09:20 - 09:40 | DM1814Mining AI-Assisted Course Design Workflows at Production Scale Aleksandr Volkov, Taras Pustovoy, Dmitriy Istomin |
| 09:40 - 09:50 | Discussion of Session 1 |
| 09:50 - 10:10 | Coffee break |
| 10:10 - 11:30 | Session 2 – Multimodal Understanding, Retrieval and Representation Mining Session chair: A/Prof. Hien Dinh Nguyen |
| 10:10 - 10:30 | S02201Context-Anchored Transformer with Masked Feature Modeling for Group Emotion Recognition Nha Tran, Minh Nguyen, Vi Huynh, Dung Dang, Hung Nguyen, Tai Huynh, Hien Nguyen |
| 10:30 - 10:50 | S02204SentiV: A Benchmark for Low-Resource Vietnamese Speech–Text Emotion Understanding Duc Dat Pham, Ngoc Tram Huynh Thi, Minh Anh Vo Ngoc, Nhan Le Thanh Pham, Tien Le Anh, Phuoc-Hai Huynh, Tan Duy Le, Kha Tu Huynh |
| 10:50 - 11:10 | S02202A Comparative Study of Similarity and Distance Metrics for Re-ranking in Retrieval-Augmented Generation (RAG) H.-A.-Khoa Ngo, K.-Duy Vo, A.-Duy Nguyen, A.-Khoi Ngo-H. |
| 11:10 - 11:30 | S02207Energy-Distance Multi-Fidelity Pruning for Hyperparameter Optimization in Deep Representation Mining Thuy Tran, Nghia Phan, Hiep Xuan Huynh |
| 11:30 - 11:50 | Open discussion Directions for world models and autonomous intelligence in data mining |
| 11:50 - 12:00 | Closing Announcement of AMI 2027 (22–24 April 2027, Vietnam) |
| 12:00 - 13:00 | Lunch (Grand Ballroom 2) |
Workshop Organizers
Prof. Hiep Xuan Huynh, Ph.D. (HDR)
Full Professor of Computer Science, leading the CTU Leading Research Team on Automation, AI, IT and Digital Transformation (CTU-AIMED). Research spans KDD, AI, IoT, and energy-based models. Published 227+ papers.
Prof. Fabrice Guillet, Ph.D. (HDR)
Full Professor of Computer Science, member of the DUKe research team. Focuses on KDD, interestingness measures, and knowledge visualization. Founding member & past president of the EGC Association.
Prof. Ryutaro Ichise, Ph.D
Professor at Institute of Science Tokyo. Research includes machine learning, knowledge graphs, semantic web, data mining, and reinforcement learning. Former Associate Professor at NII, Japan.
Tentative Program Committee
PC composition ensures broad geographic diversity (Europe, Asia, Africa, Americas) and expertise across applied intelligence systems.