Most stroke survivors face ongoing upper limb motor dysfunction. Brain-computer interface (BCI) rehabilitation technologies aim to help, but comparing different rehabilitation approaches is difficult without multi-paradigm EEG datasets from the same individuals. This study created an EEG dataset that includes various rehabilitation paradigms for the same subjects. Researchers recruited 28 healthy participants and collected EEG data through six upper limb rehabilitation paradigms. Each paradigm involved two or three actions, such as grasping and releasing with one or both hands. The dataset contains both raw EEG signals and preprocessed versions, including bandpass filtering and artifact removal. This comprehensive resource will assist in studying the neural mechanisms of different rehabilitation paradigms and contribute to developing more effective rehabilitation strategies.