CV
Education
Shandong University, Jinan, China
Sep. 2022 - Jul. 2023
Shandong University, Qingdao, China
Jul. 2023 - Now
- GPA: 88.4/100
- Rank: 6/20
Xintai No.1 Middle School, Taian, China
Secondary Education, Jun. 2022
- Within the top 1%
Experience
iLearn Qingdao, Shandong University
Research Assistant
Jul. 2023 - Now- Under the guidance of Associate Professor Xuemeng Song (宋雪萌)
- Have a certain understanding of multimodal large models.
- Currently working on applying large models to retrieval.
- Research on Composed Image Retrieval aims to allow users to retrieve target images by referencing images and modification text.
- Very familiar with a series of CIR models, which has also given a good understanding of multimodality.
人机智能中心
Intern
May. 2024 - Now- Serving as an assistant in the research of Pickball sports robots.
- Participated in explainable Go AI, aiming to enable models to explain every step of Go. Proposed the core algorithm, which was adopted by the team.
- Under the guidance of Professor Dong Xuan (宣东), an IEEE Fellow.
Projects
MLLM as a Universal Image Retriever
- We introduce an automated pipeline designed to create detailed multimodal composition datasets and present a large-scale dataset, FiGMaQ, to advance research in fine-grained context learning with MLLMs (Multimodal Large Language Models).
- We propose a two-phase fine-tuning strategy that enhances the MLLM’s capabilities in complex context understanding and query-target alignment, improving its performance across a wide range of image retrieval tasks. With distinct fine-tuning goals, this method also reduces the computational load.
- We performed comprehensive evaluations on seven diverse image retrieval datasets, demonstrating the exceptional performance of our approach. Our method achieved state-of-the-art results across several challenging image retrieval tasks in a zero-shot, single-checkpoint setting, even when using a more compact MLLM architecture.
Pseudo Triplet Guided Few-shot Composed Image Retrieval
Tools: JavaScript, Python, MySQL
GitHub- Proposed a two-stage pseudo triplet guided few-shot CIR scheme. The scheme incorporates a pseudo-triplet-based CIR pretraining stage with an attentive masking and captioning-based pseudo triplet generation method.
- Introduced a pseudo modification text-based sample challenging score estimation method and a top range-based random sampling strategy.
- First to highlight the quality of selected training triplets, maximizing the benefit of few annotated triplets.
- Plug-and-play and compatible with various CIR models, achieving improvements of up to 22.2% across datasets.
Explainable Go AI
Tools: Python, JavaScript
GitHub- Used MLLM and KataGo to align Go AI features with MLLM for game commentary generation.
- The project is currently applying for national funding for further research.
- Collaborated with a robotics company to develop an automatic Go-playing robot.
Lightweight Chatbot: Implemented with JS, Python, and MySQL
Tools: JavaScript, Python, MySQL
GitHub- Developed a lightweight chatbot for basic conversational interactions.
- Implemented a Python backend and JavaScript frontend for a seamless user experience.
- Managed conversation data using MySQL.
- Applied NLP techniques to improve response accuracy.
AI Intelligent Gomoku: Based on Java and Monte Carlo Tree Search Algorithm
Tools: Java, Monte Carlo Tree Search Algorithm
- Developed an AI Gomoku game for human-computer interaction.
- Enhanced AI decision-making using the Monte Carlo Tree Search algorithm.
- Created a user-friendly interface and optimized algorithm performance.
Smart Car: Based on Raspberry Pi, Supporting Remote Control and Automatic Line Tracking
Tools: Raspberry Pi, Python
Video- Developed a smart car with remote control and automatic line-tracking capabilities.
- Programmed car movements using Python and integrated sensors for autonomous navigation.
Mechanical Structure Analysis: Processing Structural Mechanics Problems Using Intel OneAPI
Tools: Intel OneAPI, SYCL
GitHub- Developed a program to solve complex mechanical structure problems.
- Used SYCL for parallel computing to accelerate data processing.
- Created simulation models to analyze stress conditions of structures.
Patents and Publications
[On Submession.3] Bohan Hou, Haoqiang Lin, Xuemeng Song, Haokun Wen, Meng Liu, Yupeng Hu and Xiangyu Zhao (2025). Fine-grained modeling enhancement for MLLM for Complex image retrieval . [On Submession.2] Xuemeng Song, Haoqiang Lin, Haokun Wen, Bohan Hou, Mingzhu Xu and Liqiang Nie (2024). A Comprehensive Survey on Composed lmage Retrieval. [Arxiv] [On Submession.1] Bohan Hou, Haoqiang Lin, Haokun Wen, Meng Liu, and Xuemeng Song (2024). Pseudo-triplet Guided Few-shot Composed Image Retrieval. [Arxiv]
Skills
- Programming Languages: Python, C++, C, Java, JavaScript
- English: CET-6 Score: 571
- Data Science & Machine Learning: Torch, Transformers, and other LLM platforms
- Specialized Areas: Information retrieval, Debug
- Mathematical & Statistical Tools: Mathematical analysis, linear algebra, mathematical statistics, probability theory
- Research Skills: Paper reading, experimentation, paper writing, presentation
Honors and Awards
- Second Prize in National Olympiad in Informatics (NOIP 2020) Advanced Group, Chinese Computer Federation (CCF), Nov. 2020
- Second Prize in Shandong Province High School Mathematics Competition, Shandong Provincial Mathematical Society, Oct. 2020
- Best Speed Award and Best Functionality Award in the First Intelligent Vehicle Competition, Taishan College, 2023
- Third Prize in Shandong Province Electronic Design Competition, Shandong Provincial Department of Education, 2023
- Shandong University Outstanding Student Scholarship, 2023
- Shandong University Innovation and Entrepreneurship Scholarship, 2023
- Shandong University Outstanding Student Scholarship, 2024
- Second Prize in National Undergraduate Mathematics Competition, Chinese Mathematical Society, 2024
- Second Prize in Asia-Pacific Mathematical Modeling Competition, Asia-Pacific Mathematical Modeling Competition Committee, 2024
- National First Prize in the “Shu Wei Cup” Mathematical Modeling Competition, “Shu Wei Cup” Organizing Committee, 2024
- Seventh Place in the National Finals of the “Shu Wei Cup” Mathematics Competition, “Shu Wei Cup” Organizing Committee, 2024
Additional Information
Interests: Sports, Photography, Go, Violin