Data Selection & Synthesis
Selecting and synthesizing targeted training data from model errors and capability gaps for data-efficient LMM post-training.
- Data selection
- Data synthesis
- LMM post-training
- Continual pre-training
Multimodal learning · NLP · Data-efficient Post-training
Data-Efficient Post-Training for Multimodal Foundation Models
I am a final-year Ph.D. candidate in Computer Science at UC Davis, advised by Prof. Lifu Huang. My research focuses on data-efficient post-training of multimodal foundation models.
I study data selection, data synthesis, and data-centric analysis for post-training and continual pre-training to improve long-horizon planning and multimodal reasoning in VLMs and LLMs.
March 2027 Expected graduation · Open to Research Scientist and Applied Scientist opportunities.
- ACL · EMNLP · EACL · SIGIR · AAAI
- Publication venues
- 300+
- Citations on Scholar
- 3
- Research internships
01 / Research
Research Interests
My primary research interests are in Natural Language Processing and Multimodal Learning, with a focus on Data-Efficient Post-Training of (Multimodal) Large Language Models.
Multimodal Learning
Efficiently adapting large multimodal models to emerging problems and diverse downstream tasks.
- Vision-language models
- Multimodal reasoning
- Grounding
Agentic Long-horizon Planning
Building agentic models that plan, reason, and adapt across diverse long-horizon tasks and environments.
- Agentic AI
- Multimodal planning
- Long-horizon reasoning
02 / Background
Experience & Education
Experience
- 2026
Applied Scientist Intern
Oracle
Self-Evolving Skill Optimization for Disambiguation over Knowledge Base
- 2025
Applied Scientist Intern
Amazon Alexa AI
LLM agent chain-of-thought optimization for multi-turn tool calling using SFT and RL.
- 2024
Applied Scientist Intern
Amazon Alexa AI
LLM inference latency optimization through adaptive sparsity.
- 2016–20
Software Engineer
WeBank Co., Ltd.
Education
- 2022–27
Ph.D. in Computer Science
University of California, Davis
Advisor: Prof. Lifu Huang
- 2020–22
M.S. in Computer Science & Engineering
University at Buffalo, SUNY
Advisor: Prof. Changyou Chen
- 2012–16
B.S. in Computer Science & Technology
Jilin University
Recognition
Research awards
- SIGIR 2023 Best Paper Honorable Mention
- University at Buffalo CSE Best M.S. Research Award
- Jilin University Third Prize Scholarship, 2013 & 2014
03 / Publications
Selected Publications
2025
Error-driven Data-efficient Large Multimodal Model Tuning
A data-centric approach that uses model errors to identify high-value examples for more efficient multimodal post-training.
2026
How Do Large Language Models Learn Concepts During Continual Pre-Training?
A learning-dynamics study of how large language models acquire new concepts as training continues.
2026
Self-Evolving Learning for Long-Horizon Multimodal Planning
Adaptive data synthesis and self-evolving learning for long-horizon planning across language and vision.
2023
End-to-End Multimodal Fact-Checking and Explanation Generation: A Challenging Dataset and Models
MOCHEG pairs multimodal evidence retrieval and claim verification with natural-language explanations.
2024
2024
More work5 works +
2025
2025
2023
2023
2019
04 / Updates
Selected news
How Do Large Language Models Learn Concepts During Continual Pre-Training? was accepted to EMNLP 2026.
Joined Oracle as an Applied Scientist Intern, working on Self-Evolving Skill Optimization.
Error-driven Data-efficient Large Multimodal Model Tuning was accepted to ACL 2025.
Joined Amazon Alexa AI as an Applied Scientist Intern to work on User-Facing Multi-turn LLMs Tool Use.
Joined Amazon Alexa AI as an Applied Scientist Intern to work on LLMs Inference Latency Optimization.
AMELI was accepted as an EACL 2024 Oral; FedIOD was accepted to AAAI 2024.
Received a Best Paper Honorable Mention at SIGIR 2023 for multimodal fact-checking and explanation generation.
05 / Contact
Let’s exchange ideas.
I welcome conversations about multimodal learning, data-efficient adaptation, continual learning, and agentic AI.