One of the central issues in Learning from Observation (LfO) 2.0, as discussed in the preceding chapters, involves the recognition of human demonstrations through the use of intermediate representations known as task models. Exploring intermediate representations across various task domains has been a common endeavor within LfO. The model-based design of LfO is inspired by the hypothesis that humans perceive the external world through pre-constructed models or templates [5, 169]. In the following chapters, we will examine different intermediate representations within LfO across various task domains. However, before delving into these, we will briefly assess the underlying hypothesis here.