Xinyue Ma

I build efficient systems for large language models — long-context serving, KV cache reconfiguration, and continual learning on the edge. Advised by Professor Myeongjae Jeon, Graduate School of Artificial Intelligence, OMNIA at POSTECH.

Email Curriculum Vitae ↗

Research Interests

Efficient Systems for LLMs
Continual Learning

Publications

VLDB 2026 ↗

ORBITFLOW: SLO-Aware Long-Context LLM Serving with Fine-Grained KV Cache Reconfiguration

Xinyue Ma*, Heelim Hong*, Jongseob Lee, Seoyeong Choy, Woo-Yeon Lee, Taegeon Um, Myeongjae Jeon  *Equal distribution

PDF ↗
NeurIPS 2025 ↗

REP: Resource-Efficient Prompting for Rehearsal-Free Continual Learning

Sungho Jeon, Xinyue Ma, Kwang In Kim, Myeongjae Jeon

PDF ↗
ACM MobiCom 2023 ↗

Cost-effective On-device Continual Learning over Memory Hierarchy with Miro

Xinyue Ma, Suyeon Jeong, Minjia Zhang, Di Wang, Jonghyun Choi, Myeongjae Jeon

PDF ↗ Slides Code

Experience

2025.10.12 – 2024.10.13
Web Chair
SIGOPS APSys 2025
2024.9 – 2024.12
Research Intern
Microsoft Research, Redmond — RiSE Group
Effective KV cache reuse for retrieval-augmented generation
2023.11 – 2024.2
Research Intern
Microsoft Research Asia — Intelligent Cloud & Edge Group
2024 – Present
Teaching Assistant
POSTECH
AI Systems (2025 Fall)
2022 – 2024
Teaching Assistant
UNIST
Computer Architecture (2023 Fall)
Parallel Computing (2022 Fall)
Advanced Programming (2022 Spring)