cv
Feel free to download my complete CV by clicking the Envelope button on the right.
General Information
| Name | Hao Yin |
| Label | Master's Student & Artificial Intelligence Scientist |
| yinhnavi@mail.ustc.edu.cn | |
| Phone | (+86) 18051050608 |
| Summary | Master's student in Artificial Intelligence & Data Science at USTC. My research focuses on enhancing the perception and reasoning capabilities of multimodal large language models. |
Education
-
2022.09 - Present M.Sc. in Data Science (Statistics)
University of Science and Technology of China, Hefei, China - School of Artificial Intelligence and Data Science
- GPA: 3.96/4.30 (2/31)
- Advised by Zilei Wang
-
2018.09 - 2022.06 B.Sc. in Applied Mathematics
China University of Mining and Technology, Xuzhou, China - School of Mathematics
- GPA: 4.47/5.00 (2/185)
- Outstanding Graduate
Experience
-
2020.02 - 2020.07 International Exchange Student
Australian National University, Canberra, Australia - MATH1005 Discrete Mathematical Models - High Distinction
- MATH2222 Introduction to Mathematical Thinking: Problem-Solving and Proofs - High Distinction
- MATH3511 Scientific Computing - High Distinction
-
2025.09 - 2025.12 Research Intern
Tencent Technology, Beijing, China - Research on enhancing the image captioning capabilities of MLLMs through reinforcement learning strategies.
- Developed a compact MLLM using supervised fine-tuning and reinforcement learning to generate highly precise, context-aware, and structured image captions, advancing real-world visual understanding.
- Proposed a co-evolutionary adversarial framework for MLLM image captioning that jointly optimizes captioning and question-generation models in a self-reinforcing loop, improving descriptive fidelity and completeness.
-
2026.01 - 2026.05 Research Intern
Xiaomi Technology, Beijing, China - Research on enhancing reasoning capabilities of video foundation models through post-training strategies.
- Proposed a tool-augmented MLLM reasoning framework that enables introspective reasoning across both visual and textual modalities, significantly improving long-form video understanding.
- Developed Video-OPD, an efficient post-training framework for temporal video grounding that converts sparse rewards into dense step-wise signals, accelerating convergence while outperforming existing GRPO methods.
Honors and Awards
-
2019 - National Scholarship (Top 1%), awarded at China University of Mining and Technology
- First Prize, National Undergraduate Mathematics Competition
-
2020 - National Scholarship (Top 1%), awarded at China University of Mining and Technology
-
2021 - Silver Medalist (Top 4.9%) — Cassava Leaf Disease Classification, Kaggle Competition
- Silver Medalist (Top 3.6%) - Jane Street Market Prediction, Kaggle Competition
-
2022 - Excellent Graduate, China University of Mining and Technology
Professional Services
- Conference Reviewer: ICLR (2025), CVPR (2025), ICLR (2026)