Xingda Lyu (David)

Hi! I am Xingda Lyu, a rising senior at the University of Washington pursuing dual B.S. degrees in Statistics and Informatics, and minor in Philosophy.

I conduct research with the UW NLP Group and am advised by Prof. Chirag Shah at the UW InfoSeeking Lab. I am also a summer research intern at the UIUC CS Data Mining Group, advised by Prof. Jiawei Han.

My current research focuses on three connected directions:

  • Human-AI collaboration:designing proactive, personalized agents that model users as evolving knowers and calibrate support and initiative over open-ended, long-horizon interactions.
  • Epistemic AI: grounding proactive agents in users' structural ignorance and dynamic epistemic states, regulating initiative toward epistemic partnership.
  • AI for Science: building agentic systems with dynamic retrieval and self-evolving memory for scientific reasoning.

Outside of research, I play competitive badminton and speak English, Mandarin, and French, with English and Mandarin as my native languages.

News

2026.09: Attending the Human-AI Complementarity Workshop (CMU AI-SDM) in Pittsburgh, PA.
2026.07: Attending the ICML 2026 conference in Seoul.
2026.06: Attending the CVPR 2026 conference in Denver, Colorado.
2026.06: Started as a summer research intern at the UIUC Data Mining Group, advised by Prof. Jiawei Han.
2026.05: Our paper Knowing Isn't Understanding was accepted to ICML 2026.
2026.04: Our paper PROS was accepted to the CVPR HiGen Workshop.

Publications

Preprint Overview of source-grounded problem discovery and user-governed repair for scientific posters

Beyond Instruction-Driven Editing: Source-Grounded Problem Discovery with User-Governed Repair for Scientific Posters

Xingda Lyu, Honglin Lu, Xinye Luo, and Shiqi Yang.

Preprint, 2026 · CVPR HiGen Workshop

ICML ’26 Conceptual preview for the ICML 2026 paper

Position: Knowing Isn't Understanding: Re-grounding Generative Proactivity with Epistemic and Behavioral Insight

Kirandeep Kaur, Xingda Lyu, and Chirag Shah.

In Proceedings of the International Conference on Machine Learning (ICML), 2026.

CDS ’25 Preview for the P-RAG paper

P-RAG: Prompt-Enhanced Parametric RAG with LoRA and Selective CoT for Biomedical and Multi-Hop QA

Xingda Lyu, Gongfu Lyu, Zitai Yan, and Yuxin Jiang.

In Proceedings of the International Conference on Computing and Data Science (CDS), 2025.

Workshops

CHIIR ’26 Preview for the CHIIR proactive agent presentation

Decision-Making Under Unknown Unknowns in Proactive Agentic Systems

Xingda Lyu.

ACM CHIIR Workshop on Human-Centered Proactive & Personalized Agents, 2026.

ARD ’26 Preview for the PROPER framework presentation

ProPer: Proactivity-Driven Personalized Agents for Knowledge Gap Navigation

Kirandeep Kaur, Vinayak Gupta, Xingda Lyu, Aditya Gupta, and Chirag Shah.

Amazon Research Day, AI Co-Scientist, 2026.

Research

Jun 2026-present

Research Intern, UIUC Data Mining Group

I work on AI for science, focusing on retrieval-augmented generation for retrosynthesis and drug discovery. I build and evaluate LLM/retrieval workflows for chemical reasoning and scientific decision support.

Advised by Prof. Jiawei Han.

Nov 2025-present

Research Assistant, UW InfoSeeking Lab

I study epistemic AI agents for information seeking and human-AI collaboration. My work focuses on agents that model users' structural ignorance, navigate knowledge gaps, and build stronger epistemic partnerships under uncertainty.

Advised by Prof. Chirag Shah, with PhD mentor Kirandeep Kaur.

Apr 2026-present

Research Collaborator, UW NLP Group

I study which short- and long-horizon action-effect-goal relationships actually drive an LLM's advice on consequential decisions that unfold over time. Using controlled interventions on relational structure, I test whether a model's recommendations are grounded in the relationships it articulates or driven by simpler heuristics.

In collaboration with Stella Li.

Experience

Amazon

Data Science Intern

2025.06 - 2025.08 | Seattle, WA

Baynovation

Data Science Intern

2024.07 - 2024.09 | San Jose, CA