Interaction consists of verbal exchange as well as multimodal social signals. Multimodality is essential in communicating fruitfully by sending implicit and explicit information. When conversing, multimodal signals are transferred back and forth between interlocutors. Via this transfer, interlocutors constantly adapt their behaviors to those of their interlocutors. Virtual agents, which look and act like humans, should also consider the multimodal and adaptive aspect of the interaction they build with the users. In this chapter, we elaborate on adaptive behavior evaluation methods, multimodal adaptive behavior generation modeling, and the effect of a virtual agent providing real-time adaptive behaviors for cognitive behavioral therapy (CBT).