Action- and Language-Conditioned Video Assessment for Embodied Control

Hwanhee Kim, Jaehyun Jang, Seungmin Cha +3

Vision-based embodied agents executing multi-step natural language instructions require feedback mechanisms that assess task progress over complete trajectories. Conventional appro…