About
A concise overview of my current research, academic background, and training.
Education
Rutgers University
September 2023 – Present
Ph.D. in Computer Science
Specialization: Large Language Models, Computer Vision
University of Science and Technology of China
September 2020 – June 2023
Master of Engineering, Control Science and Engineering
Specialization: Computer Vision, Object Detection, CNN Architecture
Southwest University
September 2016 – June 2020
Bachelor of Engineering, Automation
Specialization: Signals & Systems
Experience
TikTok Inc.
May 2026 – Present
Research Scientist Intern
San Jose, CA
- Built a self-evolving agent framework over a retrievable asset pool of memories, skills and tools, maintained at scale by retrieval, deduplication, revision and consolidation.
- Formalized evolution as a coverage-versus-compression trade-off, with a stopping criterion on marginal return and optimal granularity set by retriever quality.
- Characterized decision-making in LLM memory: unconditional erasure makes collapse structural, so repairs narrow an asset's boundary instead of deleting it.
Publications
Selected publications in multimodal language modeling, agents, reinforcement learning, and computer vision.
MLLM & LLM
LED: LLM Enhanced Open-Vocabulary Object Detection without Human Curated Data Generation
CVPR 2026 Website
Latent Reward Steering: An Adaptive Inference-Time Framework that Implicitly Promotes Cognitive Behaviors in Reasoning LLMs
EMNLP 2026
Multi-Agent
M3-Bench: Multi-Modal, Multi-Hop, Multi-Threaded Tool-Using MLLM Agent Benchmark
arXiv, 2025 Website
Agent Skill Evaluation and Evolution: Frameworks and Benchmarks
EMNLP 2026
Two Heads are Better Than One: Test-time Scaling of Multi-agent Collaborative Reasoning
EMNLP 2026
Act More, Decide Less: Skill-Guided Adaptive Action Chunking for Long-Horizon LLM Agents
EMNLP 2026
RL
Computer Vision & Medical Imaging
ES-VP: Energy-Shaped Dynamic Low-Rank Visual Prompting for Efficient Model Adaptation
ECCV 2026 Workshop (LIMIT)
A Multimodal Spatio-Temporal GCN Model with Enhancements for Isolated Sign Recognition
LREC-COLING 2024
A Review of Convolutional Neural Network Architectures and Their Optimizations
Artificial Intelligence Review IF 9.588/Q1
Diffusion Models for Sign Language Video Anonymization
LREC-COLING 2024
Other Collaborative Work
UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation
NeurIPS 2026
Earlier Work
Modeling & Data Analysis
A Multistory Building Evacuation Model Based on Multiple-Factor Analysis
Advances in Civil Engineering IF 1.924/Q3
Signals & Systems
The Excitation and Detection of Lamb Waves in a Droplet-Loaded Plate Using Air-Coupled Ultrasonic Transducers
Measurement IF 5.131/Q1
Quasi-Dispersion of Air-Coupled Ultrasonic Signal for Angle-Dependent Reception
Measurement IF 5.131/Q1
Multiple Reflective Signal Reception in Gas Flow Measurement Using Air-Coupled Leaky Lamb Waves
Measurement IF 5.131/Q1
Teaching & Talks
Courses and presentations that reflect my academic service and communication experience.
Computer Security
Teaching Assistant, Rutgers University (Spring 2025)
New Jersey, USA
Systems Programming
Teaching Assistant, Rutgers University (Fall 2023, Spring 2024)
New Jersey, USA
Neural Networks and Deep Learning
Teaching Assistant, University of Science and Technology of China (Aug 2021)
Hefei, Anhui, China
Characteristics of Lamb Waves and Their Leakage Waves
Presenter, ICCAR International Academic Conference (May 2019)
Beijing, China
Skills & Honors
Research interests, technical background, and selected honors.