Zhong Ye (叶重)

Zhong Ye (叶重)

Guangdong University of Technology

I’m Zhong Ye, a senior undergraduate student at Guangdong University of Technology. I’ve secured admission to the master’s program at the same university, majoring in Artificial Intelligence.

My research interests lie in the practical application of deep learning through mathematical modeling and implementation, specifically in Test-Time Adaptation (TTA), Continual Learning (CL), and AI for Science (Bioinformatics).

Publications

AAAI 2026
MT-TTA architecture

Multi-Task Test-time Adaptation via Gradient Consensus and Plasticity Constraint

Zhong Ye, Yu Hu (Corresponding Author), Zhenguo Yang

Project Description

  • We have open-sourced what is, to our knowledge, the first Test-Time Adaptation project for multi-task learning. We implemented multi-task versions of various classic TTA methods, including but not limited to Tent and EATA.
arXiv 2026
Architecture-driven Shift overview

Architecture-driven Shift: towards a lightweight selector for capturing the trends of logit shift

Zhong Ye, Yu Hu (Corresponding Author), Ruilin Tang

Project Description

  • We propose a theoretical framework, Architecture-driven Shift (ADS), enabling logit-shift estimation without a full training process in transfer and continual learning (CL) scenarios. Empirical validations show that ADS captures the tendency of logit shift well, and that the ADS-based selector is useful for reliable CL model selection.

Honors and Awards

  • 2024.12 National Scholarship (China’s highest honor for undergraduate students)
  • 2024.11 National Second Prize, China Undergraduate Mathematical Contest in Modeling
  • 2024.08 National First Prize, 2024 RAICOM National Finals (Team Leader)

Education

  • 2022.09 - 2026.06 (now), Undergraduate student at Guangdong University of Technology.

Internships

  • 2025.09 - 2026.02, Conducted optimization of few-shot speech recognition algorithms for out-of-vocabulary (OOV) circumstances, reducing the Error Rate (ER) by 8% on proprietary enterprise datasets. Shenzhen NeoAI Future Technology Co., Ltd.