Frequency Switching Mechanism for Parameter-Efficient Multi-Task Learning
Shih-Wen Liu, Yen-Chang Chen, Wei-Ta Chu, Fu-En Yang, Yu-Chiang Frank Wang · 2026
Read Frequency Switching Mechanism for Parameter-Efficient Multi-Task Learning by Shih-Wen Liu, Yen-Chang Chen, Wei-Ta Chu, Fu-En Yang, Yu-Chiang Frank Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-21111.txt.
Description
Liu and co-authors present Free Sinewich, a parameter-efficient method for multi-task fine tuning. Most efficient fine tuning methods adapt a model to one task at a time. Their layer mixes low-rank factors and convolutional priors into one kernel. A sinusoidal transform then modulates the kernel to make task specific weights at near zero cost. A small Clock Net produces bounded frequencies that keep the modulation stable. On dense prediction benchmarks, the method beats single task fine tuning while using few trainable parameters.
Catalog record
| CREATOR | Shih-Wen Liu, Yen-Chang Chen, Wei-Ta Chu, Fu-En Yang, Yu-Chiang Frank Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46905 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.21111