🗨 About Me
Hi, I'm Boyi Li. I am pursuing my dual B.S. degree in Computer Engineering at University of Illinois Urbana-Champaign and Zhejiang University.I'm currently a research intern in the Rehg Lab at UIUC, deeply honored to be mentored by Yifan Shen, Xu Cao, and Prof. James M. Rehg. Previously, I was very fortunate to be a research assistant in CVNext Lab at Zhejiang University, mentored by Zhonghan Zhao and Prof. Gaoang Wang. I am grateful to them for their help and support in my early research journey and development.
Research Interests
My research interest stems from a simple question: how do humans acquire knowledge from limited observations, form abstractions, and progressively learn about the world? My answer is visual cognition. Human vision is remarkable not because it is born complete, but because it is learned. Starting from vague blurs, infants learn to parse geometry, materials, and dynamics simply by interacting with the physical world. I think visual cognition is an intelligent solution to large-scale pattern recognition, integrating action and perception, and adaptive decision-making in complex environments. I wonder and would like to explore how human-like visual cognition mechanisms can make models implicitly simulate physical dynamics and future states and enhance models’ abilities in visual reasoning.My research explores this phenomenon through three interconnected themes.
- Cognition-Inspired Multimodal Learning: Investigating human-like visual cognitive mechanisms (e.g., mind's eye) to enable multimodal models to reason over physical dynamics and abstract concepts beyond superficial pattern matching.
- Multi-Agent Coordination & Communication: Exploring how multiple embodied agents, each with limited observations, can interact and communicate effectively to achieve collective reasoning and adaptive decision-making in complex environments.
- Developmental Modeling & AI for Pediatrics: Bridging computational models with real-world developmental data (e.g., pediatric behavioral and pose analysis), using insights from child development to inspire AI while developing clinical tools to assess and support early human growth.
Future Plans
I am currently seeking a PhD position starting in 2027 Fall, mainly foucus on the field of cognition-inspired multimodel learning. If my background aligns with your research, please contact with me, and I would be grateful to be considered and have a meeting with you if possible. I am also happy to discuss relevant potential research fit and collaboration opportunities.My Resume
News
- Aug. 2026 Our paper, ChronoVision: Temporal Reasoning via Latent State Reconstruction, is accepted by EMNLP 2026 Main Conference.
- Jul. 2026 I am very pleased to be named to the Dean’s List at UIUC for my academic performance during the Spring 2026 semester.
- Jun. 2026 Our paper, Decoding Children's Gait Behavior, is accepted by ECCV 2026.
- Jun. 2026 My personal citation reached 100! It's a meaningful milestone for me. Thanks to all my mentors for their help and support!
- May. 2026 The 1st AI Children Challenge is successfully held and all evaluations are completed! Thanks to all the teams for their support and participation!
- Mar. 2026 Our paper, The 1st AI Children Challenge, is accepted by the proceeding track of CVPR 2026 Workshop on Computer Vision for Children (CV4CHL).
- Jan. 2026 It is a great honor to serve as one of the co-organizers of CVPR 2026 Workshop on Computer Vision for Children (CV4CHL).
- Jul. 2025 Our paper, Adaptive Graph Pruning for Multi-Agent Communication, is accepted by ECAI 2025 and selected as a spotlight paper.
- Jul. 2024 Our paper, See and think: Embodied agent in virtual environment, is accepted by ECCV 2024.
📝 Selected Publications
* Equal contribution. ♡ Project lead. ✉ corresponding/co-corresponding author.
Also see Publications Page and Google Scholar for full publications.
-
Decoding Children's Gait Behavior
Yifan Shen*, Boyi Li*, Meihuan Huang*, Yuanzhe Liu*, Xu Cao*,♡, Jinyang Jin, Zhengyuan Li, Anglin Liu, Junho Kim, Jingyuan Zhu, Lan Fangzhou, Jianguo Cao, Jintai Chen, Ismini Lourentzou, James M. Rehg✉
The first and largest repository of multi-camera gait videos of children with developmental motor conditions, containing both clinically relevant assessments (EVGS and GVS) and rich frame-level annotations. The first comprehensive assessment of the ability of SoTA open and closed VLMs to decode the subtle signs of gait abnormalities from video, and find that out-of-the-box models are ineffective for this task. A novel video analysis method that is the current SoTA for the automated inference of EVGS and GVS scores from children's videos.ECCV 2026 -
Toward Cognitive Supersensing in Multimodal Large Language Model
Boyi Li*, Yifan Shen*,♡, Yuanzhe Liu*, Yifan Xu, Jiateng Liu, Xinzhuo Li, Zhengyuan Li, Jingyuan Zhu, Yunhan Zhong, Fangzhou Lan, Jianguo Cao, James M. Rehg, Heng Ji, Ismini Lourentzou✉, Xu Cao✉
A novel training paradigm that endows MLLMs with human-like visual imagery capabilities by integrating a Latent Visual Imagery Prediction (LVIP) head that jointly learns sequences of visual cognitive latent embeddings and aligns them with the answer, thereby forming vision-based internal reasoning chains. To evaluate the cognitive capabilities of MLLMs, we present CogSense-Bench, a comprehensive visual question answering (VQA) benchmark assessing five cognitive dimensions.arXiv Preprint -
Adaptive Graph Pruning for Multi-Agent Communication
Boyi Li*, Zhonghan Zhao*,♡, Der-Horng Lee✉, Gaoang Wang✉
A novel task-adaptive multi-agent collaboration framework that jointly optimizes agent quantity (hard-pruning) and communication topology (soft-pruning), dynamically constructing optimized communication topologies tailored specifically to individual tasks.ECAI 2025 -
See and think: Embodied agent in virtual environment
Zhonghan Zhao*, Wenhao Chai*,♡, Xuan Wang*, Boyi Li, Shengyu Hao, Shidong Cao, Tian Ye, Jenq-Neng Hwang, Gaoang Wang✉
A comprehensive and visionary embodied agent in the Minecraft virtual environment comprises three key components: vision perception, language instruction, and code action.ECCV 2024
🌎 Service
Reviewer:
- Conferences: KDD 2026 (AI4Science Track), KDD 2027 (AI4Science Track).
- Workshops: CVPR 2026 @ CV4CHL, CVPR 2026 @ MMRAgI, ICLR 2026 @ ES-Reasoning.