A core challenge for intelligent mobile systems lies in the inherent conflict between the growing user demand for powerful AIs and the severe resource constraints at edge. To address this, my recent works design and implement cost-effective LLM agents for edge scenarios. I use LLM post-training techniques (e.g., supervised fine-tuning, knowledge distillation, and model alignment) to boost the performance of lightweight models for specific edge applications. I am also actively exploring distributed LLM inference powered by MoE architecture. My research enables the aggregation of fragmented and underutilized computational resources across massive distributed IoT devices for supporting large-scale LLM applciations at edge.

Application-Specific Agents:

• Yang, X., Du, Y.#, Chen, K#., Liew, S. C.#, …, Heng, P. A.# "IndusGCC: A Data Benchmark and Evaluation Framework for GUI-Based General Computer Control in Industrial Automation." NeurIPS 2025. [OpenReview]
• Chen, K., Du, Y.#, Li, J., Cao, H., Guo, M., Dang, X., Li, L., Qiu, J., Chen, G. #, Heng, P. A. "ChemMiner: A Large Language Model Agent System for Chemical Data Mining." ICCV 2025. [Proceedings]
• Wang, L.*, Du, Y.*, Long, X., Liu, X., Chen, K., Liew, S. C., " Cellular-X: An LLM-empowered Cellular Agent for Efficient Base Station Operations." ACM Mobisys 2025. [ACM Library]

LLM Post-train:

• Wang, L.*, Du, Y.*, Lin, J., Chen, K., Liew, S. C. "Rephrase and Contrast: Fine-Tuning Language Models for Enhanced Understanding of Communication and Computer Networks." IEEE ICNC 2025. [IEEE Xplore]
• Zhang, Y., Chen, X., Chen, K., Du, Y.#, Dang, X., …, Heng, P. A., "The Dual-use Dilemma in LLMs: Do Empowering Ethical Capacities Make a Degraded Utility?" Major Revision at Nature Communication. [Arxiv]

Distributed MoE Inference:

• Wang, L.*, Du, Y.*#, Pan, Y., Liew, S. C.#, Liu, Y., Chen, K. "OD-MoE: On-Demand Expert Loading for Cacheless Edge-Distributed MoE Inference", Under Review at OSDI’26. [Arxiv]

Ensuring highly reliable and real-time operating links is critical for wireless-connected edge systems, particularly in mission-critical applications such as industrial automation and remote surgery. My research delivers reliable and low-latency communications through full-stack networking designs. At the lower layers (i.e., PHY and MAC), I developed ultra-reliable low-latency communication (URLLC) systems through advanced signal processing optimization and scheduling algorithms. At the application layer, I designed task-oriented semantic communication systems powered by LLMs, which ensure the successful execution of planned actions even in the presence of communication errors.

Task-Oriented Communication:

• Zhang, F.*, Du, Y.*, Xiang, Y., Liu, X., Liew, S. C. " SA-OOSC: A Multimodal LLM-Distilled Semantic Communication Framework Leveraging Contextual Understanding for Coding Efficacy", Accepted by IEEE ICNC 2026. [Arxiv]
• Zhang, F.*, Du, Y.*, Chen, K., Shao, Y., Liew, S. C., "Out-of-Distribution in Image Semantic Communication: A Solution with Multimodal Large Language Models." IEEE Transactions on Machine Learning in Communications and Networking, 2025. [IEEE Xplore]
• Zhang, F.*, Du, Y.*, Chen, K., Shao, Y., Liew, S. C., "Addressing Out-of-Distribution Challenges in Image Semantic Communication Systems with Multi-modal Large Language Models." IEEE WiOpt 2024. [IEEE Xplore]

URLLC:

• Du, Y., Liew, S. C., "Reliable Packet Detection for Random Access Networks: Analysis, Benchmark, and Optimization." IEEE Transactions on Vehicular Technology, 2025. [IEEE Xplore]
• Du, Y., Hao, L., Lei, Y., "Nonlinear Multi-Carrier System with Signal Clipping: Measurement, Analysis, and Optimization." IEEE Systems Journal, 2024. [IEEE Xplore]
• Du, Y., Liew, S. C., Shao, Y., "Efficient FFT Computation in IFDMA transceivers." IEEE Transactions on Wireless Communications, 2023. [IEEE Xplore]
• Du, Y., Hao, L., Lei, Y., "SER Optimization in OFDM-IM Systems with Nonlinear Power Amplifiers." IEEE Transactions on Vehicular Technology, 2023. [IEEE Xplore]

Another important line of research of mine is to develop real-world intelligent edge systems with practical impacts for both the research community and commercial R&D. My recent works explore Industrial 4.0 systems, with an emphasis on soft manufacturing – where advanced AI and wireless communication technologies are integrated to enable unmanned system automation and flexible product line reconfiguration. I am now building a fully autonomous self-driving laboratory for automated pharmaceutical manufacturing in collaboration with chemical experts. Meanwhile, I am actively investigating the synergy of AI and wireless communication to advance remote healthcare applications in the upcoming 6G era.

Given the interdisciplinary nature, my works in this line of research are conducted in collaboration with domain experts from Harvard, National University of Singapore (NUS), Zhejiang University (ZJU), and the CSE Department at CUHK.

Industrial IoT Automation (with a drug manufacturing use case):

• Cui, H.*, Du, Y.*, Yang, Q., Shao, Y., Liew, S. C., "LLMind: Orchestrating AI and IoT with LLMs for complex task execution." IEEE Communications Magazine, 2025. [IEEE Xplore]
• Du, Y., Yang, Q., Wang, L., Lin, J., Cui, H., Liew, S. C. "LLMind 2.0: Distributed IoT Automation with Natural Language M2M Communication and Lightweight LLM Agents," Under Review at IEEE Journal on Selected Areas in Communications. [Arxiv]
• Xu, S.*, Zhang, L.*, Du, Y.*, Liew, S. C. "A Hybrid TDMA/CSMA Protocol for Time-Sensitive Traffic in Robot Applications", Accepted by IEEE ICRA 2026. [Arxiv]
• Chen, K., Li, J., Wang, K., Du, Y.#, …, Heng, P. A., "Chemist-X: Large Language Model-empowered Agent for Reaction Condition Recommendation in Chemical Synthesis.", Major Revision at Computational Chemistry. [Arxiv]

Remote Healthcare:

• Dang, X., Chen, K., Su, X., …, Long X., Du, Y., Zitnik, M., Heng, P. A." KnowGuard: Knowledge-Driven Abstention for Multi-Round Clinical Reasoning", accepted by ICLR 2026. [OpenReview]
• Du, Y., Chen, K., Zhan, Y., Low, C.H., Islam, M., Guo, Z., Jin, Y., Chen, G., Heng, P. A. " LMT++: Adaptively Collaborating LLMs with Multi-specialized Teachers for Continual VQA in Robotic Surgical Videos." IEEE Transactions on Medical Imaging, 2025. [IEEE Xplore]
• Chen, K.*, Du, Y.*, You, T., Islam, M., Guo, Z., Jin, Y., Chen, G., Heng, P. A. "LLM-Assisted Multi-Teacher Continual Learning for Visual Question Answering in Robotic Surgery." IEEE ICRA 2024. [IEEE Xplore]