Evolutionary Excellence: Bridging the Gap in Multilingual AI

Zhou, Yu

Chen, Xianwei

Tan, Kay Chen

Wu, Xingyu

Faculty of Computer and Mathematical Sciences

Department of Data Science and Artificial Intelligence

Faced with the challenge of building affordable, multilingual AI, PhD students Zhou Yu and Chen Xianwei turned to evolutionary computation. Competing in the global FLAME 2024 challenge, they developed a resource-efficient ""bi-level"" approach, combining parameter merging with evolutionary architecture search to fuse models across different languages and tasks. By implementing a ""Glue Layer"" with knowledge distillation, they stabilized cross-model connections while ensuring cross-language fairness. Despite intense time pressure and computational limits, the team’s commitment to reproducible research and modular design turned a complex technical challenge into a career-defining success, bridging the gap between evolutionary computation and modern AI.

Learning-to-learn

Project Management and Teamwork

Resourcefulness and Adaptability to New Contexts

Adaptability and Flexibility

Strategic Planning

First Place, IEEE CIS FLAME Technical Challenge 2024 [International competition]

2026-08

PhD

PhD

Evolutionary computation

Natural language processing (Computer science)

Multilingual computing

Computer architecture

All rights reserved

English

Feature Story

open access