CV / ResumeAbout Me
Hello! I am Lutao Yan (Neal, 晏璐涛), an MPhil student in Data Science and Analysis at the Hong Kong University of Science and Technology (Guangzhou). I received my BEng in Data Science and Big Data Technology from South China University of Technology.
My research focuses on multimodal learning, chart understanding, visual analytics, information retrieval, and LLM applications. I am particularly interested in building reliable models and data systems that help people search, understand, and reason over real-world charts.
Education
MPhil, Data Science and Analysis
BEng, Data Science and Big Data Technology, School of Future Technology
Publications & Research
These selected publications form a connected research path in multimodal chart understanding—from evaluating low-level reasoning and constructing reusable training data to semantic retrieval and robust visual alignment. An asterisk (*) denotes equal contribution.
ChartInsights: Evaluating Multimodal Large Language Models for Low-Level Chart Question Answering
Introduces a low-level ChartQA benchmark covering 10 tasks and 7 chart types, evaluates 19 multimodal models, and proposes Chain-of-Charts to improve chart data question answering.
ChartCards: A Chart-Metadata Generation Framework for Multi-Task Chart Understanding
Builds an automated chart-metadata generation framework and the 85K-chart MetaChart dataset, enabling shared training data across retrieval, question answering, and other chart tasks.
Boosting Text-to-Chart Retrieval through Training with Synthesized Semantic Insights
Introduces the real-world BI benchmark CRBench and a hierarchical semantic-insight training pipeline for ChartFinder, improving precise text-to-chart retrieval NDCG@10 to 66.9%.
ChartAlign: Instance-Level Visual Alignment for Robust Chart Understanding in MLLMs
Constructs visually diverse but semantically equivalent ChartPairs and aligns image encoders at the instance level, improving robust multimodal reasoning on unlabeled and artistic charts.
Experience
Industry Experience
Supported the commercialization of AIGC products through data analysis and risk-control infrastructure, including detection of abusive content and anomalous behavior.
Built data and sampling pipelines for mixed-ranking CTR models using real search logs, user profiles, behavioral history, and query intent. Improved personalized ranking for feed-like search scenarios through data-flow and model-flow optimization.
Research Experience
Advised by Prof. Yuyu Luo and Prof. Weikai Yang. Conducted four visual-analytics studies on multimodal chart understanding, data construction, and retrieval.
Designed LLM evaluation systems and benchmarks, and reviewed generated text and annotations against quality standards.
Advised by Prof. Fangxin Wang. Explored sparse models, including mixture-of-experts architectures, for federated learning and edge intelligence.
Advised by Prof. Ye Liu and Prof. Jin Xu. Studied cross-domain lie detection and model generalization across data domains.
Advised by Prabhu Natarajan. Implemented traffic-sign recognition and image-enhancement methods for complex visual environments.
Awards
CNY 10,000 per month
CNY 10,000 per month
CNY 5,000 scholarship
CNY 5,000 scholarship
CNY 5,000 scholarship
Third Prize
Successful Participant
Top 10%
Excellence Award
Skills
Python, SQL, Java, C/C++, Pandas, NumPy, MySQL, Tableau
PyTorch, Transformers, RAG, agentic workflows, function calling, MCP, prompt engineering
OpenAI Codex, Claude Code, Git/GitHub, Linux, VS Code, Google Cloud Platform, LaTeX