Efficient LLM Inference
KV-cache reuse, semantic cache distillation, state transfer, selective recomputation, and bandwidth-aware serving.
AI Systems Research
I work on efficient LLM systems, semantic state transfer, multi-candidate reasoning, and security-oriented machine learning.
Efficient LLM inference, reasoning reliability, and trustworthy machine learning.
KV-cache reuse, semantic cache distillation, state transfer, selective recomputation, and bandwidth-aware serving.
Candidate construction, fixed-verifier reranking, coverage-conversion gaps, and answer-mode compatibility.
Backdoor attacks, frequency-domain robustness, model behavior under perturbation, and security evaluation.
Reproducible experiments, GPU resource management, and AI-assisted research workflows.
Five accepted papers: four at CCF A conferences and one at a CCF B conference.
Forty-third International Conference on Machine Learning (ICML), 2026. (CCF-A)
Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026. (CCF-A)
Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026. (CCF-A)
34th ACM International Conference on Multimedia (ACM MM), 2026. (CCF-A)
IEEE International Conference on Web Services (ICWS), 2026. (CCF-B)
Tools and workflows for efficient inference, reliable evaluation, and reproducible research.
KV-cache reuse, semantic cache distillation, selective patching, and ongoing vLLM integration for efficient inference.
Controlled comparisons and ablations for inference performance, candidate selection, and model safety.
Linux/SSH, Conda, GPU experiment deployment and resource management, and Codex/MCP-assisted research workflows.
Open to research collaboration.
Zhuhai, China