#multimodal agent

'M3-Agent is a new multimodal framework that creates entity-centric episodic and semantic long-term memories from continuous video and audio, enabling improved multi-turn reasoning and stateful behavior over time.'