Large Language Models (LLMs) spanning language, vision, audio, and other modalities are rapidly transforming the AI landscape, enabling a wide range of downstream applications. As demand for more capable models continues to rise, both model scale and training data volume have expanded substantially. Training, fine-tuning, and serving such models increasingly rely on large-scale high-performance computing (HPC) systems and remain highly resource- and time-intensive.

Data compression has emerged as a promising means of mitigating communication and data-movement overhead in distributed and parallel environments for modern AI and big-data workloads. Because data movement across the Internet, inter-node networks, and system interconnects has become a major determinant of both runtime and energy consumption, efficient mechanisms for data transfer and analysis are increasingly critical.

This workshop addresses key research challenges in reducing data-movement and communication costs for large-scale AI and big-data applications, including model training, fine-tuning, inference, and emerging LLM-based agent and multi-agent systems.

Topics of interest include but are not limited to:

• Data Compression Methods

  ° Compression Techniques for Structured and Unstructured Scientific Data
  ° Image, Video, and Multimedia Data Compression
  ° Time-series Data Compression
  ° Textual Data Compression (Natural Language, Logs)
  ° Quantization and Data Reduction
  ° Predictive Coding and Transform-based Compression
  ° Dictionary-based and Entropy-based Compression
  ° Tensor Decomposition and Low-rank Approximations
  ° Compression-aware Data Mining and Machine Learning
  ° Compression for Accelerating Data Analytics

• Applying Data Compression in AI-Related Applications and Systems

  ° Large-Scale AI Model Training
  ° Large-Scale AI Model Fine-Tuning
  ° Large-Scale AI Inference/Serving
  ° LLMs-Based Agent and Multi-Agent System Designing
  ° Data Compression for Communication Reduction
  ° Data Compression to Reduce Memory and Storage Overhead

• Hardware Co-Design for Applying Data Compression in Emerging AI Applications, Big Data Applications, and Quantum Computing

  ° GPUs
  ° FPGAs
  ° Quantum Computing Platforms
  ° CXL: Compute Express Link
  ° PIM: Process in Memory
  ° RISC-V
  ° ARM

• Papers should be submitted electronically on the ICPP submission system:

https://ssl.linklings.net/conferences/icpp/

• Paper submission must be in ACM format:

https://www.acm.org/publications/proceedings-template

• DC4AI will accept full papers (limited to 10 pages including references) and short papers (6 pages, including references and appendix).

• Submitted papers will be evaluated by at least 3 reviewers based on technical merits.

• DC4AI encourages submissions to provide artifact description & evaluation.

• Accepted papers that are presented in the workshop will be published in the ACM Digital Library.