where: Center for Machine Learning Research, Peking University, independently
Time: Dec. 2023 – Present
- Pain Point: In multi-task learning, the scheduling order of datasets during fine-tuning of large models is animportant but overlooked point, so I proposed a corresponding solution
- Pioneered the application of the ”Chain of Thought” concept in Curriculum Learning, with a particular focus on Multi-task Learning and data dependencies among datasets
- Explored the research workflow extensively, demonstrating proficiency in independent ideation, experimentaldesign, and iterative refinement processes
- Acquired proficiency in PyTorch, Hugging Face, and Adapterhub, enhancing expertise in deep learning frameworksand model repositories