Wenxuan Song

Integration of diverse robot policies can be streamlined from hours to minutes, revolutionizing how we deploy and evaluate robotic systems.

Robust-WAM achieves superior out-of-distribution generalization in robot control by seamlessly integrating semantic foresight into action predictions while leveraging extensive VGM pretraining.

MobileWAM achieves superior mobile manipulation performance by seamlessly integrating foresight into action planning, outpacing state-of-the-art methods.

Trajectory guidance boosts mobile manipulation success rates by over 20%, transforming how robots interpret and execute complex tasks.

Achieving a 57.6% success rate on RoboCasa365, Xiaomi-Robotics-1 sets a new standard for vision-language-action models in real-world robotic manipulation.

MLLMs falter in fine-grained interpersonal reasoning, but integrating visual cues and social roles can dramatically boost their performance.

MLLMs struggle to juggle proactive tasks and reactive queries in dynamic video streams, but a simple agentic framework can significantly improve their coordination without any training.

Ditch the clunky architectures: a single diffusion model can now handle vision, language, and robot control to achieve SOTA manipulation performance.

A practical VLA model, LLaVA-VLA, achieves strong generalization and versatility on a new benchmark, CEBench, while running on consumer-grade GPUs, eliminating the need for costly pre-training.

By aligning latent representations with multiple visual foundation models, FRAPPE offers a more scalable and data-efficient way to imbue generalist robotic policies with robust world-awareness.