Although the unity and diversity model of executive functions (EFs) has been replicated, there are some studies questioning the validity of the EFs construct. This debate can be partially resolved by directly combining the brain activity pattern in different executive control processes. Previous univariate activation studies have suggested that the neural substrates of different EFs (eg, updating, inhibiting, and shifting) involve common and distinct brain regions. However, the underlying multivariate neural representation of EFs in terms of unity and diversity is still elusive. Here, we employed the n-back task, stop signal task, and category switching task to investigate the characteristic of the neural representation in the three EF domains. At the global level, multivoxel pattern analysis revealed that a three-way classifier built with global activation pattern successfully distinguished the three EF tasks. At the local level, although most overlapping activations exhibit lower neural representational similarity, the inferior frontal junction showed similar neural representation across the three EFs, which was further confirmed by searchlight analysis that additionally revealed other similar representational regions were located in the presupplementary motor area extend to dorsal midcingulate cortex. In addition, using machine learning-based predictive framework, the resting-state functional networks built with the representational regions of EFs predicted intellectual abilities to some extent in a large independent sample. These findings suggest that different EFs are characterized by dissociable global neural representation but also share similar local neural representation, which contributes to understanding the neural correlates of the unity and diversity of EFs from an integrated framework.(PsycInfo Database Record (c) 2022 APA, all rights reserved)