Duke University
2024.08 – PresentPh.D. Student, Electrical and Computer Engineering
Advised by Yiran Chen and Hai Li
About Me
Hello! My name is Qinsi Wang. I am a third-year PhD student in the CEI lab of the Department of Electrical and Computer Engineering at Duke University. I am fortunate to be advised by Prof. Yiran Chen and Prof. Hai "Helen" Li.
Before that, I conducted research at the University of Science and Technology of China and the National University of Singapore, where I was fortunate to be mentored by Prof. Lin Shao. I received my undergraduate degree in Electronic Science and Technology from Huazhong University of Science and Technology.
Research Interests
My recent research focuses on two directions:
01
Post-training and Agentic Self-improvement
How can a model keep improving with little or no human data?
- Angles Don't Lie (NeurIPS 2025, Spotlight): Models know what data they need and should select training data themselves.
- Vision-Zero (ICLR 2026): Models can improve reasoning ability by playing strategic games against copies of themselves.
- From RLVR to RLSVR (COLM 2026): Building verifiable rewards into the environment makes even unverifiable tasks trainable, with no external verifier or reward model.
02
Multimodal Reasoning
How can multimodal LLMs reason well enough to help people in the real world?
- InteractiveWorld (arXiv 2026): An environment that simulates real-world interaction between humans and MLLM assistants, for evaluating and training perception, memory, proactivity, and communication of MLLMs.
- Seeing is Solving (ICML 2026): Visual matching lets MLLMs filter their own RL training data to improve reasoning.
News
- May 2026I’ll be joining
Meta Reality Labs in Burlingame as a summer research intern in 2026! Looking forward to meeting more outstanding people. 🎉😊✨
- Apr 2026CoreInfer and See is Solving have been accepted by ICML 2026! At the same time, I am honored to have been selected as a Gold Reviewer.
- Jan 2026Vision-Zero has been accepted by ICLR 2026! Code is released at project page.
- Apr 2025Angles Don't Lie has been accepted by NeurIPS 2025 (Spotlight)! Code is released at project page.
- May 2025I’ll be joining Adobe in San Jose as a summer research intern in 2025! Always open to coffee chats ☕😊 — feel free to reach out!
- Apr 2025A little personal milestone: I’ve had first-author papers accepted at all three major AI conferences — ICML, ICLR, and NeurIPS! Cheers🎉!
- Apr 2025CoreMatching has been accepted by ICML 2025. Code is released at project page.
- Feb 2025Dobi-SVD has been accepted by ICLR 2025. Dobi-SVD is a novel LLM compression solution on low-cost computation devices! Visit the project page for more information.
- Oct 2024CoreInfer has been uploaded to arXiv. CoreInfer achieves a 10.33x speedup on an NVIDIA Titan XP without sacrificing performance! Visit the project page for more information.
- Mar 2024I am excited to announce that I will join the Department of Electrical and Computer Engineering at Duke University as a PhD student in Fall 2024! Looking forward to my PhD life!
- Sep 2023MathNAS has been accepted by NeurIPS 2023. See project page for more information.
- Apr 2023DGL has been accepted by the IEEE Transaction on Mobile Computing (CCF A). Visit the project code for more information.
- Jul 2022I received my B.S. from Huazhong University of Science and Technology (HUST) and received the title of Outstanding Graduate!
- Sep 2021I received the China National Scholarship (0.2%)! This is the highest award given by the Ministry of Education of China for undergraduates.
Selected Publications
Post Training / Agentic Self-improvement
Multimodal Reasoning
Efficiency LLM/VLMs
Education
Huazhong University of Science and Technology
2018.09 – 2022.06Bachelor's Degree, Electronic Science and Technology
Internship
Meta
2026.05 – 2026.12Research Intern · Manager: Qian Qi, Wei Wen
Key Achievements: Developed egocentric AI assistant to enhance the intelligence and contextual awareness of Meta AI glasses. Investigated how embodied assistants perceive, reason about, and interact with the physical world.
Adobe
2025.05 – 2026.02Research Intern · Mentor: Wentian Zhao
Key Achievements: Developed a self-improving agent system to enhance autonomous learning and performance. Published three first-author papers and received Adobe’s Gift Fund in recognition of outstanding intern performance.
Hobby
I love photography and travel, here are some of my favorite photography works and wonderful life moments!
