ERIA 26 Introduction

ERIA26, short for Efficient Running of Intelligent Agents 2026, is a special track of ChinaSoft 2026 dedicated to key technologies for the efficient operation of intelligent agents.

As large models rapidly evolve from content generation to task execution, intelligent agents are emerging as a core paradigm for next-generation intelligent systems, driving artificial intelligence from isolated capabilities toward system-level deployment.

To address complex real-world requirements, intelligent agents must continuously perceive virtual and physical environments, understand intentions, formulate plans, invoke tools, and execute tasks. They must also maintain stable operation, support self-evolution, and perform fine-grained adaptation through closed-loop feedback.

In cloud-edge-device collaborative and cyber-physical environments, the efficient operation of intelligent agent systems faces multiple challenges, including heterogeneous multi-source data, heterogeneous software and hardware platforms, limited resources on edge and end devices, and stringent security and reliability requirements in open environments.

Research on system software, system optimization, hardware-software co-design, reliability and security, and energy-efficient computing for intelligent agents therefore has significant academic value and practical importance.

• Special Issue Title
   
  Key Technologies for the Efficient Operation of Intelligent Agents
   
• Organizing Committee (Guest Editors)
   
  • Rui Han, Beijing Institute of Technology
  • Yuzhong Sun, Institute of Computing Technology, Chinese Academy of Sciences
  • Zhengwei Qi, Shanghai Jiao Tong University
  • Jianxin Li, Beihang University

1. Scope and Topics of Interest

This special track focuses on key technologies for the efficient operation of intelligent agents. It aims to bring together the latest research advances and practical applications in system software, artificial intelligence, cloud-edge-device collaboration, embodied intelligence, intelligent terminals, and the Industrial Internet, thereby promoting academic exchange and industrial deployment in related fields.

Topics of interest include, but are not limited to, the following:

• 1. Hardware Adaptation and Acceleration of Large Models for Intelligent Agents.
  Hardware-aware adaptation, hardware-software co-design, and acceleration techniques for the efficient execution of large models on GPUs, FPGAs, ASICs, and other computing platforms.

• 2. Compression and Lightweight Deployment of Large Models for Intelligent Agents.
  Model compression and lightweight deployment techniques, including pruning, quantization, knowledge distillation, and related methods that balance model accuracy and inference efficiency.

• 3. Memory Management and Storage Optimization of Large Models for Intelligent Agents.
  Memory management, data movement, and storage optimization strategies for improving computational and operational efficiency in resource-constrained environments.

• 4. Distributed Multi-Agent Computing and Device-Cloud Collaborative Inference.
  Distributed computing and collaborative inference techniques that integrate computing resources across end devices, edge platforms, and cloud systems to achieve efficient inference and optimized resource allocation.

• 5. Adaptive Inference and Dynamic Optimization for Intelligent Agents.
  Adaptive inference and dynamic optimization algorithms that adjust model configurations according to changing environments and task requirements, achieving an effective balance between performance and resource consumption.

• 6. Semantic-Aware and Isolation-Enabled Runtime Infrastructure for Intelligent Agents.
  Underlying architectures, runtime environments, and management mechanisms that support the concurrent execution of multiple intelligent agents while improving execution precision, efficiency, isolation, and robustness.

• 7. Security and Robustness Enhancement of Large Models for Intelligent Agents.
  Security defense, reliability assurance, and robustness enhancement techniques that ensure system reliability and behavioral controllability in complex and uncertain environments.

• 8. Energy-Efficient and Sustainable Intelligent Agent Systems.
  Energy-aware and low-power optimization techniques for reducing energy consumption and supporting the efficient and sustainable operation of intelligent agents and their underlying large models.

• 9. Industrial Case Studies and Practical Experience.
  Industry-oriented case studies and practical experience in intelligent agent system optimization, including applications in embodied intelligence, autonomous driving, intelligent Internet of Things systems, and other representative domains.

2. Submission Guidelines

2.1 Paper Types

The following types of manuscripts are welcome:

1. Original research papers
2. High-quality survey and review papers
    

2.2 Submission Requirements

1. Manuscripts should be submitted through the EasyChair submission system at:

   https://easychair.org/my/conference?conf=eria26

   For detailed submission requirements and manuscript formatting guidelines, please refer to:

   http://cjc.ict.ac.cn/wltg/zgjz.htm

2. The submitted manuscript must present the authors’own research results. All data must be authentic and reliable, and the manuscript must not have been previously published. The work of others must be properly cited, and there must be no disputes concerning authorship. The manuscript should provide a complete presentation of research findings with original academic insights and make a meaningful contribution to the development of the relevant academic field.

3. The manuscript should be written fluently and clearly, with good readability and without logical or grammatical errors. Reviewers should be able to read and understand the manuscript without difficulty.

3. Important Dates

Phase I: Submission to ChinaSoft 2026

• Paper Submission Deadline

  September 1, 2026

  Authors should select the corresponding track when submitting their manuscripts.

• Notification of Review Results

  September 15, 2026

• Revised Manuscript Deadline

  October 15, 2026

  The track committee will review the revised manuscripts and evaluate the revisions made by the authors.

Phase II: Conference Presentation and Revised Submission

ChinaSoft 2026 Conference

November 2026, Tianjin, China

At least one author of each paper selected during Phase I must attend the conference, present the paper in person, and respond to questions from the committee members and the audience.

Papers that are not presented on site will be withdrawn from further consideration.

Revised Manuscript Submission Deadline

The revised manuscript should be submitted within one week after the oral presentation at ChinaSoft 2026.

Authors should revise their manuscripts based on the comments and suggestions received from committee members and the audience during the conference presentation.

 

Phase III: Recommendation for Journal Review

Notification of Recommendation

December 15, 2026

The track committee will evaluate the conference papers and notify the authors whose papers have been selected for recommendation.

Selected outstanding papers will be recommended to the Chinese Journal of Computers. The journal will conduct an additional internal peer-review process before making the final publication decision.

4. Contact

For questions concerning the special issue and manuscript submissions, please contact:

Yuzhong Sun
Email: yuzhongsun@ict.ac.cn

5. Invitation to Submit

We sincerely invite experts, scholars, researchers, and practitioners working at the forefront of related fields to submit their latest research findings and practical experience.

Through this special issue track, we aim to advance the theoretical development and industrial application of intelligent agent technologies and promote the sharing and practical deployment of relevant research outcomes.