Frequency-Decoupled Cross-Modal Knowledge Distillation

Junhong Liu, Yuan Zhang, Tao Huang, Haozhao Wang, Wenchao Xu, Shanghang Zhang, Renyu Yang

Keywords: Cross-Modal Knowledge Distillation (CMKD), Frequency-Domain Feature Decoupling, Feature Space Alignment

TL;DR: We propose a novel CMKD method using frequency-domain feature decoupling, achieving superior results over traditional KD and state-of-the-art CMKD methods.

Abstract:

Knowledge distillation (KD) has proven highly effective for compressing large models and enhancing the performance of smaller ones. However, its effectiveness diminishes in cross-modal scenarios, such as vision-to-language distillation, where inconsistencies in representation across modalities lead to difficult knowledge transfer. To address this challenge, we propose frequency-decoupled cross-modal knowledge distillation, a method designed to decouple and balance knowledge transfer across modalities by leveraging frequency-domain features. We observe that low-frequency features tend to capture modality-agnostic, generalizable information, while high-frequency features are more modality-specific. Accordingly, we apply distinct losses to these features: enforcing strong alignment in the low-frequency domain and introducing relaxed alignment for high-frequency features. Additionally, we propose a scale consistency loss to address distributional shifts between modalities, and employ a shared classifier to unify feature spaces. Extensive experiments across multiple benchmark datasets show that our method substantially outperforms traditional KD and state-of-the-art cross-modal KD approaches.

Primary Area: applications to computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 465