publications

Publications

2026

  1. EMNLP 2026
    AgentKV: Phase-Aware KV Eviction for Agentic LLMs
    Taowen Tony Liu, Jeffrey TH Wong, Can Xiao, and 3 more authors
    Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, 2026
    Paper
  2. ICML 2026
    A3: an Analytical Low-Rank Approximation Framework for Attention
    Jeffrey TH Wong, Cheng Zhang, Xinye Cao, and 4 more authors
    arXiv preprint arXiv:2505.12942, 2026
    Paper Code Website
  3. UCRL @ ICLR’26
    On the Existence and Behaviour of Secondary Attention Sinks
    Jeffrey TH Wong, Cheng Zhang, Louis Mahon, and 3 more authors
    arXiv preprint arXiv:2512.22213, 2026
    Paper Code
  4. MAIGAI @ ICLR’26
    Team of Thoughts: Efficient Test-time Scaling of Agentic Systems through Orchestrated Tool Calling
    Jeffrey TH Wong, Zixi Zhang, Junyi Liu, and 1 more author
    arXiv preprint arXiv:2602.16485, 2026
    Oral Paper Code
  5. PA @ ICML’26
    Rethinking Scaffolding in LLM Tutors: The Interactional Mismatch Between Benchmarks and Real-World Deployments
    A Neagu, Jeffrey TH Wong, M Messer, and 2 more authors
    In ICML 2026 Workshop on Pluralistic Alignment, 2026
  6. ISCA 2026
    Combating the Memory Walls: Optimization Pathways for Long-Context Agentic LLM Inference
    Haoran Wu, Can Xiao, Jiayi Nie, and 15 more authors
    2026
    Paper

2025

  1. ICLR 2025
    QERA: an Analytical Framework for Quantization Error Reconstruction
    Cheng Zhang, Jeffrey TH Wong, Can Xiao, and 2 more authors
    The Twelfth International Conference on Learning Representations, 2025
    Paper