Proceedings

Parameter-Efficient Fine-Tuning with Circulant and Diagonal Vectors

Xinyu Ding, Lexuan Chen, Siyu Liao, Zhongfeng Wang

Download BibTeX Cite

Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence. AI4Tech: AI Enabling Technologies. Pages 9185-9193.

https://doi.org/10.24963/ijcai.2025/1021

Keywords

  • Advanced AI4Tech: Deep AI4Tech
  • Advanced AI4Tech: Generative and LLMs-driven AI4Tech
  • Advanced AI4Tech: Creative AI4Tech

Abstract

Foundation models have achieved tremendous success in different domains.

However, their huge computation and storage complexity make these models difficult to fine-tune and also less applicable in practice.

Recent study shows training in Fourier domain can be an effective fine-tuning method in terms of both model performance and number of training parameters.

In this work, we propose to further reduce the complexity by the factorization through the product of interleaved circulant and diagonal matrices. In addition, we address the case of non-square fine-tuning weights by partitioning the circulant matrix into blocks.

Our method avoids the construction of weight change matrix and utilizes 1D fast Fourier transform (FFT) instead of 2D FFT.

Experimental results show that our method achieves similar or better performance across various tasks with much less floating-point operations (FLOPs) and the number of trainable parameters.

Xinyu Ding, Lexuan Chen, Siyu Liao, and Zhongfeng Wang. 2025. Parameter-Efficient Fine-Tuning with Circulant and Diagonal Vectors. In Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-25. IJCAI, 9185–9193. https://doi.org/10.24963/ijcai.2025/1021
Welcome Future Conferences AI Journal Division IJCAI Proceedings About IJCAI IJCAI Awards