Reproduce “Frequency Switching Mechanism for Parameter-E!cient Multi-Task Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shih-Wen Liu, Yen-Chang Chen, Wei-Ta Chu, Fu-En Yang, Yu-Chiang Frank Wang
Free Sinewich adapts one shared low-rank weight base for multiple dense-vision tasks. A task token sets a frequency, and an elementwise sine transforms the base into task-specific weights. On PASCAL-Context, it reports +5.39% average relative improvement over single-task fine-tuning with 6.53M trainable parameters, versus +4.24% and 7.38M for TADFormer at rank 64; this is a benchmark result, not a runtime claim.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.