Yushi SUN

I am currently a senior researcher at LightSpeed Studio, Tencent Games (Project Up, 青云计划). I obtained my PhD degree from Hong Kong University of Science and Technology (HKUST) in 2025, majored in Computer Science and Engineering. My supervisor is Prof. Lei CHEN. I am fortunate to work with Prof. Nan TANG and Dr. Xin Luna Dong during my PhD journey.

I received the B.S. degree in Computer Science and Applied Mathematics from Hong Kong University of Science and Technology (HKUST) in 2021. More details of me are introduced in my CV|简历.

Email  /  Google Scholar  /  Github  /  Linkedin

My research interest lies in the following topics: Data Management (Data Annotation, Data Integration, and Data Organization), RAG, Agent Memory, and RL. My vision is to make data usable: starting from ensuring the annotation and preparation quality, to developing effective data organization paradigm, to improving usability of data in downstream applications (e.g., RAG). Here are the selected publications. Please check my Google Scholar to get the full list of my publication. Feel free to drop me an email if you are interested in potential collaborations! :)

We propose APEX-SQL, an Agentic Text-to-SQL Framework that shifts the paradigm from passive translation to agentic exploration. Our framework employs a hypothesis-verification loop to ground model reasoning in real data. In the schema linking phase, we use logical planning to verbalize hypotheses, dual-pathway pruning to reduce the search space, and parallel data profiling to validate column roles against real data, followed by global synthesis to ensure topological connectivity. For SQL generation, we introduce a deterministic mechanism to retrieve exploration directives, allowing the agent to effectively explore data distributions, refine hypotheses, and generate semantically accurate SQLs.

[LLM Agent]

We introduce a novel PLM-LLM collaborated cross data lake column type annotation approach that significantly reduces the generalization cost of domain specifc column type annotation models.

[Data Preparation]

We proposed MedKGI, a diagnostic framework grounded in clinical practices. MedKGI integrates a medical knowledge graph (KG) to constrain reasoning to validated medical ontologies, selects questions based on information gain to maximize diagnostic efficiency, and adopts an OSCE-format structured state to maintain consistent evidence tracking across turns.

[KG]

We proposed a novel KGQA approach that utilizes a caching structure to significantly improve the QA accuracy of KG-based RAG pipeline.

[LLM, RAG, KG]

We proposed a novel KGQA approach that resolves the low recall issue of existing methods.

[LLM, RAG, KG]

We constructed a comprehensive RAG benchmark and hosted the KDD Cup competition.

[LLM, RAG, Data Organization]

We conducted an extensive evaluation of SOTA LLMs on taxonomies.

[LLM, Data Organization]

We proposed a novel cross-domain-aware worker selection with training approach for crowdsourced data labeling.

[Data Annotation]

We defined a novel named entity schema for related and sub-related table discovery and alignment for enhancing the annotation quality of column semantic types.

[Data Preparation]