Feng Yu

I am Feng Yu, a PhD student in Computer Science at the University of Exeter, supervised by Prof. Jia Hu and Prof. Geyong Min. My research focuses on continual learning and efficient adaptation for foundation models, with a particular interest in post-training methods for LLMs and LLM agents under evolving tasks, distributions, and workflows.

My recent work approaches this question from two angles: BoHA, a LoRA-style method for parameter-efficient LLM fine-tuning, and Fed-TaLoRA, a task-agnostic low-rank residual adaptation method for dynamic incremental tasks in distributed training settings.

I am open to discussing research ideas and academic collaborations. Feel free to reach out by email. My research is supported by China Scholarship Council and University of Exeter Scholarships.

News

May 18, 2025	One research paper FedTaLoRA has been released!
Nov 20, 2024	The first comprehensive survey FCL for Edge-AI has been released!
Sep 1, 2023	Started my PhD at Exeter!

Selected publications

2025

BoHA: Blockwise Hadamard Product Adaptation for Parameter-Efficient Fine-Tuning

Feng Yu , Jia Hu, and Geyong Min

arXiv preprint arXiv:2509.21637 2025

Bib HTML
@article{yu2025blockwise, title = {BoHA: Blockwise Hadamard Product Adaptation for Parameter-Efficient Fine-Tuning}, author = {Yu, Feng and Hu, Jia and Min, Geyong}, journal = {arXiv preprint arXiv:2509.21637}, year = {2025}, bibtex_show = {true}, selected = {true}, abbr = {arXiv}, html = {https://arxiv.org/abs/2509.21637} }
Task-agnostic Low-rank Residual Adaptation for Efficient Federated Continual Fine-Tuning

Feng Yu , Jia Hu, and Geyong Min

arXiv preprint arXiv:2505.12318 2025

Bib HTML
@article{yu2025efficient, title = {Task-agnostic Low-rank Residual Adaptation for Efficient Federated Continual Fine-Tuning}, author = {Yu, Feng and Hu, Jia and Min, Geyong}, journal = {arXiv preprint arXiv:2505.12318}, year = {2025}, bibtex_show = {true}, selected = {true}, abbr = {arXiv}, html = {https://arxiv.org/abs/2505.12318} }