Omni Interaction Agent Technical Report
Gander is presented as a full-duplex multimodal agent that can listen, watch, speak, and act in one streaming system.
The report says the model continuously takes video, speech, and text instead of waiting for turn-based prompts. Its design splits realtime interaction from deeper reasoning through a “Cerebellum-Brain” framework. The authors say evaluations cover conversation, omni understanding, interaction, and agentic intelligence, with internal human tests showing competitive omni interaction and strong spoken dialogue. They also say Gander, its models, code, and data are being released for research use. ArXiv · AI/CL/LG's note
The report says the model continuously takes video, speech, and text instead of waiting for turn-based prompts. Its design splits realtime interaction from deeper reasoning through a “Cerebellum-Brain” framework. The authors say evaluations cover conversation, omni understanding, interaction, and agentic intelligence, with internal human tests showing competitive omni interaction and strong spoken dialogue. They also say Gander, its models, code, and data are being released for research use. ArXiv · AI/CL/LG's note
score 6