IJCAI-ECAI 2026 · Half-day Tutorial
Principles, Structures, and Adaptive Agents:
Scaling AI under Limited Supervision
A unified framework spanning data-efficient priors, graph/relational structure as implicit supervision, and adaptive agentic systems
🎯 Data Efficiency · Graph Supervision · Adaptive Agents
⏳ Half-day (3h30m incl. break)
📖 Tutorial Overview
How can AI systems scale when direct supervision is limited? This half-day tutorial connects data-efficient priors, graph and relational structure, and adaptive agents across settings with scarce labels, weak feedback, and self-generated supervision.
🎓 Target Audience
Researchers, practitioners, and graduate students in machine learning, foundation models, graph learning, multi-agent systems, and AI for science. Basic familiarity with machine learning and neural networks is sufficient.
✨ Distinctive Contribution
A unified path from priors → structure → interaction, with practical guidance for comparing methods, evaluating reliability, and choosing limited-supervision strategies.
📌 Tutorial at a Glance:
A practical framework for building and evaluating AI systems with scarce labels, weak feedback, or self-generated supervision.
A practical framework for building and evaluating AI systems with scarce labels, weak feedback, or self-generated supervision.
⏱️ Tutorial Schedule (Half-day, includes 25-minute break)
| Time | Session / Topic | Presenter(s) |
|---|---|---|
| 09:00 - 09:05 | 🎬 Opening & Framing Supervision bottlenecks, unifying taxonomy, cross-cutting failure modes (bias amplification, confirmation loops, unstable self-improvement) View slides (PDF) | All (lead: Q. Yao) |
| 09:05 - 09:55 | 📌 Module A: Principles — Learning with Limited Labels This module explains how prior experience enables learning from a small support set. It compares optimization-, metric-, and amortization-based meta-learning with in-context learning, then shows how to formulate practical few-shot problems through molecular prediction and cold-start recommendation examples. View slides (PDF) | Yaqing Wang |
| 09:55 - 10:45 | 🔗 Module B: Structures — Relational & Graph Implicit Supervision This module shows how graph structure extracts more information from scarce labels, rollouts, and interactions. It introduces experience graphs for data collection, decision graphs for planning and action selection, and learning graphs for feedback propagation and credit assignment. View slides (PDF) | Zixing Song |
| 10:45 - 11:10 | ☕ Break & Networking | |
| 11:10 - 11:55 | 🤖 Module C: Adaptive Agents under Limited Feedback This module examines how agents adapt during interaction when feedback is sparse or delayed. It develops structured context as editable external state for retrieving and revising experience, then connects decision-level adaptation to reusable workflows, dynamic multi-agent organization, and adaptive knowledge routing. View slides (PDF) | Quanming Yao |
| 11:55 - 12:30 | 🧠 Synthesis, Discussion & Q&A Decision-oriented comparison, evaluation principles, trustworthy scaling, open research challenges, and audience discussion | All (moderated by Yao) |
※ Total duration: 3 hours 30 minutes including a 25-minute break. Each module blends core principles, representative methods, practical design trade-offs, and open questions.
👥 Presenters
Lecturer (UK Assistant Professor), University of Bristol
Zixing Song is currently a Lecturer (UK Assistant Professor) at the University of Bristol. Prior to that, he was a Postdoctoral Research Associate at the University of Cambridge. He received his Ph.D. in Computer Science and Engineering from The Chinese University of Hong Kong. He also obtained his B.Eng. in Computer Science from Southeast University in China. His research centers on data-centric graph machine learning and its application to drug design and broader AI4Science domains, driven by a vision of developing data-efficient, agentic, and trustworthy models for social good. Currently, he serves as Vice Chair of the IEEE Computational Intelligence Society (CIS) Task Force on Data-efficient Agentic Learning. He has also received several international awards, including the AAAI New Faculty Highlight, the Junior Research Fellowship at Wolfson College Cambridge, and the INNS Doctoral Dissertation Award.
Associate Professor, Beijing Institute of Mathematical Sciences and Applications (BIMSA)
Yaqing Wang is an Associate Professor at the Beijing Institute of Mathematical Sciences and Applications (BIMSA). She received her Ph.D. in Computer Science and Engineering from the Hong Kong University of Science and Technology (HKUST). Her research interests lie in machine learning and artificial intelligence, with a particular focus on data-efficient learning, including few-shot learning, in-context learning, and adaptive AI agents. She has authored over 40 research papers in leading venues such as NeurIPS, ICML, ICLR, KDD, The Web Conference, IEEE TPAMI, JMLR, and TIP. Her publications have received more than 6,000 citations. Dr. Wang serves as an Associate Editor of Neural Networks, an Editorial Board Member of Machine Learning, and an Area Chair for ACL Rolling Review. She is also an Associate Chair of the IEEE CIS NNTC Task Force on Data-Efficient Agentic Learning. She is a recipient of the Hong Kong PhD Fellowship Scheme and the Beijing Nova Program, and has been recognized among the World's Top 2% Scientists in both 2024 and 2025. She was invited to present her research at the AAAI New Faculty Highlights 2026 and the IJCAI Early Career Spotlights 2026.
Associate Professor, Tsinghua University
Quanming Yao is an Associate Professor and Ph.D. advisor in the Department of Electronic Engineering at Tsinghua University. He received his Ph.D. in Computer Science and Engineering from the Hong Kong University of Science and Technology (HKUST). Before joining Tsinghua, he was a researcher and senior scientist at 4Paradigm, where he founded and led its machine learning research team. His research spans machine learning and AI for science, with a focus on data-efficient agentic learning, few-shot and in-context learning, knowledge graphs, and biomedical networks. He has published over 120 papers, receiving more than 15,000 citations. Dr. Yao is an IET Fellow and a recipient of IEEE Computing's Top 30, the INNS Aharon Katzir Young Investigator Award, the inaugural Ant InTech Prize, Forbes China 30 Under 30, and the Google Ph.D. Fellowship. He has delivered invited early-career talks at AAAI and IJCAI. He serves as an Area Chair for ICML, NeurIPS, and ICLR; an Action Editor for IEEE TPAMI, Machine Learning, and TMLR; and a Senior Action Editor for Neural Networks. He was Tutorial Chair of IJCAI 2025 and is Program Co-chair of ADMA 2026.
📌 Together, the presenters offer an integrated view across priors, structure, and interaction.
⚖️ Responsible AI & Key Risks
The tutorial explicitly addresses risks arising under limited supervision: bias amplification from scarce labels, error propagation in pseudo-labeling/self-training, spurious correlations in graph data, and instability in self-improving agent loops. We emphasize practical safeguards: confidence calibration, selective human oversight, robust evaluation under weak supervision, and transparent reporting of failure modes and domain limits. These issues are especially critical in scientific, biomedical, and socially consequential settings — limited supervision must never be confused with reduced responsibility. Trustworthy scaling is framed as both a technical and scientific question: improving capability while preserving reliability, transparency, and appropriate human control.
📎 Reference List
- Generalizing from a Few Examples: A Survey on Few-Shot Learning. ACM Computing Surveys, 2020.
- Finetuned Language Models Are Zero-Shot Learners. ICLR, 2022.
- Why In-Context Learning Models are Good Few-Shot Learners? ICLR, 2025.
- ReAct: Synergizing Reasoning and Acting in Language Models. ICLR, 2023.
- ExpeL: LLM Agents Are Experiential Learners. AAAI, 2024.
- No Change, No Gain: Empowering Graph Neural Networks with Expected Model Change Maximization for Active Learning. NeurIPS, 2023.
- Beyond Scaling: A Survey on Data-Efficient Agentic Learning. IJCAI, 2026.
- Reflexion: Language Agents with Verbal Reinforcement Learning. NeurIPS, 2023.
- G-Memory: Tracing Hierarchical Memory for Multi-Agent Systems. NeurIPS, 2025.
- Language Model Networks: Supervision-Efficient Learning through Dense Communication. ICML, 2026.
- AFlow: Automating Agentic Workflow Generation. ICLR, 2025.
📬 Contact & Organization
For inquiries regarding the tutorial, please contact the organizers: Quanming Yao (Tsinghua University) qyaoaa@tsinghua.edu.cn. Further details and live updates will be available on the official IJCAI-ECAI 2026 conference website.
🏷️ #Limited Supervision 🏷️ #Graph SSL 🏷️ #Adaptive Agents 🏷️ #Few Shot Learning 🏷️ #Agentic AI