About 3 results
Open links in new tab
Mod-Squad: Designing Mixtures of Experts As Modular Multi-Task Learners Zitian Chen1, Yikang Shen2, Mingyu Ding3, Zhenfang Chen2, Hengshuang Zhao3, Erik Learned-Miller1, Chuang …
A. Comparison with γ-MoD Concurrent to our work, γ-MoD [33] also propose to inte-grate Mixture-of-Depths mechanism into MLLMs. In this section, we first analyze the difference between our …
A6. Task relation from different layers of Mod-Squad. In the paper, we define the similarity between tasks as the mean of the percentage of experts that they are sharing given the same …