Hao Yao
M.S. Student
Master’s student at CUIT. Research focuses on optimization of model and data heterogeneity in large-scale training of Multimodal Large Language Models (MLLM).
Education
M.E.
- PresentChengdu University of Information Technology
B.E.
- PresentChengdu University of Information Technology
Work Experience
Contestant
- PresentTencent AI Arena
Algorithm Designer
- PresentChengdu Yujunsheng Technology Co., Ltd.
Teaching Secretary
- PresentSchool of Computer Science, CUIT
Projects
SelCo: Multimodal Training Framework Based on Heterogeneous Co-location
Research
Academic Research
Role:Core Researcher
2025.12
-
Present
Proposed Selective Co-location mechanism for MLLM large-scale training heterogeneity. System includes generative module for candidate construction and enumeration filtering, plus MILP optimizer module. Results submitted to ICIC 2026 with related invention patent.
Automated Annotation and Object Detection Based on SAM 3
Practice
Internship Project
Role:Developer
2026.01
-
2026.02
Explored computer vision models for object detection tasks. Tested and deployed SAM 3 model to optimize automated data annotation workflow.
Escape (RPG Game Development)
Practice
Personal Development
Role:Game Designer & Developer
Implemented RPG core combat mechanics and character action systems using C# and Unity. Built enemy AI system with patrol and sentry modes using A* pathfinding.
Run Away (2D Two-player Adventure Game)
Practice
Personal Development
Role:Game Designer & Developer
Built game scenes and character models in Unity. Implemented two-player cooperative gameplay with elastic rope module and themed levels.
2CarRacing (3D Two-player Racing Game)
Practice
Collaborative Development
Role:Game Designer & Developer
Built core mechanics in Unity engine with vehicle physics model and motion algorithms. Implemented dual-camera real-time rendering and split-screen display.
Publications
SelCo: Efficient Distributed Multimodal LLM Training with Selective Co-location.
Proceedings of the 22nd International Conference on Intelligent Computing (ICIC 2026), Toronto, Canada,
2026.
SelCo: Efficient Distributed Multimodal LLM Training with Selective Co-location.
Proceedings of the 22nd International Conference on Intelligent Computing (ICIC 2026), Toronto, Canada,
2026.