Learning from competitive interactions is essential for improving the capabilities of both single- and multi-agent systems. For embodied agents, where physical interactions play a key role, competition becomes more complex, especially when dealing with adversarial behaviors between limbs. To explore how agents can learn from competition, this chapter proposes a competitive learning framework that enables individuals to acquire knowledge through interactive competition. Additionally, an adversarial learning framework is introduced to facilitate the extraction of knowledge from adversarial behaviors. Furthermore, a joint-optimization approach is presented to evolve agent designs and tactics in cooperative interactive confrontations, allowing agents to adapt effectively to their opponents.