Yanlin Qi (齐艳琳)

Ph.D. Candidate @ Université Paris Cité
Supervisor: Prof. Themis Palpanas (ACM Fellow, IUF Senior Fellow)

If you are interested in my research, I would be happy to connect and collaborate.

I am a Ph.D. candidate in Computer Science at Université Paris Cité. My current work focuses on algorithm-system co-design for sparse attention in large models, covering both training and inference. I am interested in sparse attention, attention routing, long-context modeling, and efficient large-model reasoning.

I am actively looking for research internship opportunities in sparse attention, large models, and efficient training and inference.

Previously, I studied Mathematics at Harbin Institute of Technology, Shenzhen, and worked on pattern mining, learned indexes, data selection, and LLM-driven optimization.

Yanlin Qi portrait
Keep moving.
Pursue what I love.
yanlinqi7@gmail.com Google Scholar GitHub X LinkedIn CV WeChat
Education
  • UP
    Université Paris Cité
    Ph.D. in Computer Science
    Dec. 2024 - Present
    Sparse attention for large models; algorithm-system co-design for training and inference.
  • HIT
    Harbin Institute of Technology, Shenzhen
    M.Sc. in Mathematics
    Sept. 2021 - Jan. 2024
  • DMU
    Dalian Maritime University
    B.Sc. in Statistics, Distinction
    Sept. 2017 - Jun. 2021
    Ranked 1st out of 59 students.
Experience
  • HK
    HKUST(GZ)
    Research Assistant with Prof. Yuyu Luo
    Oct. 2024 - Dec. 2024
    Data selection for LLM instruction tuning.
  • CityU
    City University of Hong Kong
    Research Assistant with Prof. Zhichao Lu
    Mar. 2024 - Sept. 2024
    LLM-driven evolutionary computation and automatic algorithm design.
  • HKUST
    Hong Kong University of Science and Technology
    Visiting Student with Prof. Qiyu Liu and Prof. Lei Chen
    Sept. 2023 - Mar. 2024
    AI for databases, learned indexes, and vector compression.
Honours and Awards
  • Top 10, Agentic Hackathon Paris
    2026
  • Excellent Master's Thesis, Harbin Institute of Technology
    2024/01
  • First-class Scholarship, Harbin Institute of Technology
    2021-2022
  • National Scholarship
    2018-2021
  • Excellent Student Award, Dalian Maritime University
    2021/05
  • Innovation and Entrepreneurship Training Programme Award
    2019-2020
Research Interests
  • Sparse attention for large models, covering training and inference.
  • Algorithm-system co-design for attention routing and long-context computation.
  • Approximate search, indexing, and selection algorithms for efficient model computation.
News
  • 2026. ParisKV was accepted to ICML 2026.
  • 2026. I am actively looking for research internship opportunities in sparse attention and large-model efficiency.
  • 2026. LEAD was accepted to VLDB 2026.
  • 2025. Learned-index and SegPQ papers appeared in PVLDB 2025.
Selected Publications
ParisKV Fig. 2 pipeline
ParisKV: Fast and Drift-Robust KV-Cache Retrieval for Long-Context LLMs

Yanlin Qi, Xinhang Chen, Huiqiang Jiang, Qitong Wang, Botao Peng, Themis Palpanas

ICML 2026

KV-cache retrieval for long-context LLMs; connects sparse attention, Top-k selection, and robustness under long generation.

[Paper] [Code]

LEAD framework figure
LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning

Xiaotian Lin, Yanlin Qi, Yizhang Zhu, Themis Palpanas, Chengliang Chai, Nan Tang, Yuyu Luo

VLDB 2026

[Paper]

Learned index illustration
Why Are Learned Indexes So Effective but Sometimes Ineffective?

Qiyu Liu, Siyuan Han, Yanlin Qi (co-first), Jingshu Peng, Jin Li, Longlong Lin, Lei Chen

PVLDB 2025

[Paper]

SegPQ framework figure
Not Small Enough? SegPQ: A Learned Approach to Compress Product Quantization Codebooks

Qiyu Liu, Yanlin Qi (co-first), Siyuan Han, Jingshu Peng, Jin Li, Lei Chen

PVLDB 2025

[Paper]

RecRanker training pipeline
RecRanker: Instruction Tuning Large Language Model as Ranker for Top-k Recommendation

Sichun Luo, Bowei He, Haohan Zhao, Wei Shao, Yanlin Qi, Linqi Song

ACM TOIS

[Paper]

Closed high-utility pattern mining extension figure
Mining Periodic Trends via Closed High-Utility Patterns

Yanlin Qi, Guoting Chen, Wensheng Gan

Expert Systems with Applications

[Paper]