Academic
About Me#
I am an undergraduate in Computer Science and Technology at Xi'an Jiaotong University, working on artificial intelligence. My interests include AI Infra, multimodal learning, and weakly supervised learning.
I have received the National Scholarship for two consecutive years and five national awards in total.
I am interested in improving the general capabilities of models. I share the perspective behind The Bitter Lesson: data, infrastructure, and compute set the limits of what models can achieve. I place more trust in high-quality data and efficient AI infrastructure than in elaborate algorithmic designs. I also use large models to improve my research productivity, while keeping the thinking in my own hands. Understanding a problem deeply is essential to judging, guiding, and making good use of a model's capabilities.
I enjoy exploring the unknown and pursue it as a personal interest.
Here is my academic CV, feel free to download it.
Download CV
Research Interests#
AI Infra
Model Quantization, KV Caching, and Speculative Decoding
Multimodal Learning
Multimodal Representation Learning, Text-to-Image
Weakly Supervised Learning
Learning with Noisy Label, Semi-Supervised Learning
Publications#
2026
Learning What to Trust in Multimodal Learning under Noisy Supervision
REFINE detects noisy labels in multimodal learning by combining fused and unimodal representations. It selects reliable representation spaces for each class and evaluates label consistency within them, providing cleaner supervision for more robust multimodal classification.
Representative Awards#
National Scholarship
Academic Year 2025-2026
National First Prize
Chinese Mathematics Competitions for College Students
National First Prize
National Finals, New Liberal Arts Practice and Innovation Competition for College Students
National Scholarship
Academic Year 2024-2025
National Second Prize
China College Students Engineering Practice and Innovation Ability Competition