Beyond Episodic Evaluation: Memory Architectural Bottlenecks in Sequential Embodied Question Answering

Part of

Real-Time Action Conditioned Video Generation