I am Huyu Wu(伍胡宇), a graduate student at the Institute of Computing Technology, Chinese Academy of Sciences (ICT). I received my bachelor’s degree in Artificial Intelligence from the School of Computer Science.
My research interests lie in the Multimodal Large Language Models and Dataset Distillation. You can find my main research interests on 
🔥 News
- 2025.09: 🎉🎉 One paper is accepted by Neural Networks.
- 2024.10: 🎉🎉 One paper is accepted by Applied Soft Computing.
- 2024.06: 🎉🎉 One paper is accepted by Expert Systems with Applications.
- 2024.03: 🎉🎉 One paper is accepted by IJCNN 2024
-
📝 Publications
- Wu H, Jia B, Yuan X M. LLM-Led Vision-Spectral Fusion: A Zero-Shot Approach to Temporal Fruit Image Classification[J]. Neural Networks, 2025: 108155.(First Author, CCF-B)
- Jia B, Guo Z, Huang T, Guo F, & Wu H. A generalized Lorenz system-based initialization method for deep neural networks[J]. Applied Soft Computing, 2024, 167: 112316. (Last Author, 中科院一区TOP)
- Jia B, Wu H, Guo K. Chaos theory meets deep learning: A new approach to time series forecasting[J]. Expert Systems with Applications, 2024, 255: 124533. (Co-First Author, 中科院一区TOP)
- Wu H, Jia B, Sheng G. Early-Late Dropout for DivideMix: Learning with Noisy Labels in Deep Neural Networks[C]//2024 International Joint Conference on Neural Networks (IJCNN). IEEE, 2024: 1-8. (First Author, CCF-C)
📖 Educations
- 2025.09 - Present, Institute of Computing Technology, University of Chinese Academy of Sciences
- 2021.09 - 2025.06, College of Computer Science, Artificial Intelligence, Sichuan University
- 2020.09 - 2021.06, Business School, Business Administration, Sichuan University
💻 Internships
-
2024.08 - Present, Intern of TSAIL - Tsinghua University
-
2024.04 - 2024.08, Intern of Institute of Automation Chinese Academy
- 2023.12 - 2024.04, Research Assistant of National University of Singapore - Institute of Operations Research and Analytics
- 2022.11 - 2024.08, Research Assistant of Sichuan University - Machine Intelligence Lab
📙 Projects
Diffusion Models As Dataset Distillation Priors
Dataset distillation seeks to create compact yet informative datasets, but balancing diversity, generalization, and representativeness remains challenging. We propose Diffusion As Priors (DAP), which leverages the inherent representativeness prior in diffusion models by quantifying synthetic–real data similarity in feature space with a Mercer kernel and guiding the reverse diffusion process without retraining.
When Language Overrules: Revealing Text Dominance in Multimodal Large Language Models
Multimodal Large Language Models (MLLMs) achieve strong results but suffer from text dominance, where textual inputs outweigh other modalities. We present the first systematic study of this issue across images, videos, audio, time-series, and graphs, introducing two metrics—the Modality Dominance Index (MDI) and Attention Efficiency Index (AEI)—to quantify the imbalance. Our analysis reveals its pervasive nature and key causes, and we propose a simple token compression method that rebalances attention, reducing LLaVA-7B’s MDI from 10.23 to 0.86. This work offers insights and methods for building more balanced multimodal models.
Dataset Condensation with Color Compensation
Dataset condensation faces inherent trade-offs between performance and fidelity. Current methods struggle with inefficiency (image-level selection) or semantic distortion (pixel-level optimization). We identify color’s dual role as an information carrier and semantic unit as critical. To address this, we propose DC3, which enhances color diversity via latent diffusion models after calibrated image selection. Experiments show DC3 outperforms SOTA methods across benchmarks. Notably, DC3 enables fine-tuning pre-trained diffusion models with condensed datasets without degradation (validated by FID scores), marking the first exploration of this capability.
VEX Robotics Competition
Led the school’s VEX Robotics club, responsible for programming and debugging of the robotic systems.
Attained Gold Awards at China Zone Selections, the Asia Championships, Asia Open and the World Championships in the United States during the 2018 season.
🎖 Honors and Awards
- 2023.06 Champion, 3rd Youth Campus Volleyball League of Sichuan Province.
- 2022.11 Second Prize, Asia-Pacific Undergraduate Mathematical Contest in Modeling.
- 2022.06 Third Prize, 2th Youth Campus Volleyball League of Sichuan Province.
📚 Academic Services
-
Conference Reviewer: IJCNN 2024, ACL 2024, CaLM @NeurIPS 2024, ICLR 2025, AAAI 2026
-
Journal Reviewer: IEEE Transactions on Neural Networks and Learning Systems(IEEE TNNLS)