Topics

Potential topics include, but are not limited to:

Track 1: Foundations & Theoretical Aspects of General Intelligence

• Mathematical & Logical Foundations: Information Theory, Optimization, Geometry, and Causal Inference

• Learning Theory & Cognitive Architectures: Generalization, Continual Learning, Meta-Learning, and Biologically-Inspired Designs

• World Modeling & Physics-Informed AI: Physical Reasoning, World Models, and Commonsense Knowledge Representation

• Mechanistic Interpretability & Theoretical Analysis: Expressiveness, Scaling Laws, and Representation Theory

Track 2: Large Model Architectures, Training & Post-Training

• Advanced Model Architectures: Transformers, State-Space Models (SSMs), Mixture-of-Experts (MoE), and Hybrid Designs

• Data Curation & Pre-training Methodologies: Data Synthesis, Filtering, Governance, and Pre-training Algorithms

• Post-Training & Alignment Techniques: Reinforcement Learning (RLHF/RLAIF), Direct Preference Optimization (DPO), and Instruction Tuning

• Test-Time Compute & Reasoning Scaling: Inference-Time Search, Process Supervision, and Deliberate Reasoning

Track 3: Autonomous Agents, Complex Reasoning & Embodied AI

• Complex Reasoning & Planning: Logical Deduction, Symbolic Reasoning, and Multi-Step Decision-Making

• Autonomous & Multi-Agent Systems: Tool-Augmented Agents, Memory Systems, Reflection, and Collaborative Multi-Agent Intelligence

• Multimodal Perception & Cross-Modal Learning: Vision-Language Models, Audio/Video Processing, and Spatial Intelligence

• Embodied AI & Physical World Interaction: Robotics Control, Embodied Perception, and Physical-World Grounding

Track 4: Efficient Systems, Infrastructure & Edge Computing

• Distributed Training & Infrastructure: Distributed Systems, Fault Tolerance, Parallelism Strategies, and Network Interconnects

• Hardware Acceleration & AI Co-Design: AI Chip Architectures, Hardware Acceleration, and Hardware-Software Co-Design

• Model Compression & Efficient Inference: Quantization, Pruning, Knowledge Distillation, and Sparse Attention

• Edge Intelligence & Real-Time Systems: On-Device Deployment, Edge Inference, and Low-Latency Computing

Track 5: Trustworthiness, Safety, Evaluation & Domain Applications

• Safety, Alignment, and Governance: Human Value Alignment, Red Teaming, Controllable Generation, and AI Governance

• Evaluation & Benchmarking Frameworks: Comprehensive Evaluation, Dynamic Benchmarks, and Robustness Assessment

• Knowledge-Augmented Systems: Retrieval-Augmented Generation (RAG), Dynamic Knowledge Graph Integration, and Enterprise AI

• AI for Science & Cross-Disciplinary Applications: Scientific Discovery, Healthcare, Finance, and Industrial Intelligence