Yinghao Xu

LingBot-VA 2.0 achieves few-shot generalization in complex robot manipulation tasks, outperforming traditional video generative models.

Unbounded interaction horizons and real-time responsiveness redefine the possibilities for immersive AI-driven environments.

LingBot-Video bridges the gap between digital creativity and physical actuation, achieving unprecedented efficiency in video pretraining for embodied intelligence.

Navigation models trained in Image2Sim's synthetic environments outperform traditional methods, achieving zero-shot transfer to real-world settings.

Semantic visual-action tokenization in RepWAM significantly enhances robotic manipulation performance, outperforming traditional reconstruction-based approaches.

Achieving a 93.1% improvement in training accuracy and 2x faster inference, Next Forcing redefines the efficiency of causal world modeling in video generation.

Real-time 3D scene reconstruction from streaming video is now possible with a feed-forward transformer that outperforms traditional SLAM methods.

A million videos with paired depth, camera pose, and 3D point tracks could unlock a new wave of 3D-aware video models.