Workshop V


Digital Twin-Augmented Large Language Model Agents

Workshop Organising Committee

Chair

Jiaxing Shen

Lingnan University, Hong Kong

jiaxingshen@LN.edu.hk

Co-chair

Shi You

Nanjing Tech University, China

shiyou@njtech.edu.cn

Co-chair

Yuzhi Zhou

Hong Kong Metropolitan University, Hong Kong

s1315400@live.hkmu.edu.hk

TPC Chair

Longyu Zhou

Singapore University of Technology and Design, Singapore

zhoulyfuture@outlook.com

Abstract

Large language model (LLM) agents have emerged as critical enablers for information integration, generation, and reasoning across increasingly complex application domains. To enhance their performance and situational intelligence, techniques such as prompt engineering, context augmentation, and advanced fine-tuning have attracted significant research attention. The timely and accurate provision of structured contextual and background information decisively governs the effectiveness of LLM agents. Concurrently, digital twin technology provides a promising paradigm to fulfil this requirement by continuously synchronizing user states, environmental conditions, and task progress, modeling them as expressive features, and delivering these dynamic world models to LLM agents. Despite the promising trend, the systematic coordination between digital twins and LLM agents remains uncharted, presenting a host of open challenges spanning data synchronization, semantic and token alignment, and autonomous decision-making. Inspired by the recent surge of interests, this workshop aims to investigate the envisioned intersection of LLM agents and digital twin technologies, covering the full lifecycle of data acquisition, real-time state synchronization, semantic modelling, agent learning, communication, and practical deployment. We seek original work not currently under review by any other journal, magazine, or conference.

Potential Topics

  • —  Digital twin-driven data acquisition and context modeling for LLM and agents
  • —  Real-time synchronization and state consistency mechanisms in digital twin systems
  • —  Semantic modeling, knowledge graph construction, and memory management for digital twins and LLM agents
  • —  Prompt engineering and in-context learning enhanced by digital twin feedback
  • —  Digital twin-empowered fine-tuning, alignment, and personalization of LLM agents
  • —  Digital twin aware function-calling, tool-use, and API orchestration augmentation
  • —  Digital twin empowered reasoning, planning, and model construction
  • —  Token-efficient inter-agent communication and collaboration protocols
  • —  Decentralized resource orchestration and task scheduling for digital twin and LLM agent networks
  • —  Privacy-preserving, secure, and trustworthy data sharing for agent learning via digital twins
  • —  LLM agents for intelligent optimization and management of digital twin edge networks
  • —  Benchmarking, evaluation metrics, and simulation platforms for digital twin and LLM agents
  • —  Continual learning and lifecycle adaptation of LLM agents through digital twins
  • —  Human-in-the-loop interaction, explainability, and value alignment in digital twin and LLM agent systems

TPC Members

  • Wenshuai Liu, Jiangnan University, China
  • Tao Qi, Jiangnan University, China
  • Yongkang Li, Hong Kong Metropolitan University, Hong Kong
  • Jiayuan Chen, Nanjing University of Aeronautics and Astronautics, China
  • Ruoyang Chen, Nanjing University of Aeronautics and Astronautics, China
  • Yuye Yang, Nanjing University of Aeronautics and Astronautics, China

Submission

Submission details refer to the ICAIT 2026 website: https://www.icait.org/submission.html