The behavior of traffic participants is full of uncertainties in the real world. To safely navigate through traffic, autonomous vehicles should anticipate the other vehicles’ intentions and evaluate the threats, taking seriously the ones with high probability to happen and not overreacting to the ones with low probability. Probabilistic intention and motion predictions are unavoidable to accomplish safe and high-quality decision-making and motion planning for autonomous vehicles[1]. Wrong predictions can generate conservative motions, such as unnecessary stops/yielding, or dangerous situations, like emergency brakes and collisions.