Chairs

Introduction

Foundation models, including large language models and multimodal models, have introduced substantial challenges in computation, memory capacity, communication efficiency, and system scalability. Their training and deployment increasingly rely on large-scale parallelism, distributed execution, heterogeneous accelerators, and efficient runtime management. Meanwhile, emerging workloads such as long-context inference, mixture-of-experts models, and high-concurrency online serving place additional pressure on system performance, resource utilization, and reliability. This special track aims to bring together researchers and practitioners working on parallel and distributed techniques, system architectures, scheduling mechanisms, and runtime support for scalable, efficient, and dependable foundation-model training and inference.

Topics

• Scalable training and efficient adaptation of foundation models
• Efficient inference and high-performance model serving
• Long-context processing and KV-cache management
• Mixture-of-experts and retrieval-augmented generation systems
• Resource scheduling and runtime management for foundation models
• Heterogeneous computing and hardware–software co-design
• Model compression, quantization, and memory-efficient execution
• Reliable, secure, and energy-efficient foundation-model deployment

Important Date

We are pleased to announce the important dates for PDCAT 2026. Please mark your calendars!
All deadlines are based on Anywhere on Earth (AoE), at midnight on the specified date.

• Paper Submission Deadline: 31 August, 2026 (AoE)
• Notification of Acceptance: 10 October, 2026
• Camera-Ready Submission Deadline: 31 October, 2026
• Conference Dates: 4-6 December, 2026