Si Liu

Incorporating panoramic spatial context into VLA policies boosts mobile manipulation success rates by over 20% compared to local-view methods.

Despite advances in MLLMs, they still struggle with dynamic reasoning, falling far short of human capabilities in interpreting continuous visual cues.

CR-Solver achieves over 95% success rates and millimeter-level accuracy in motion generation for continuum robots, all while leveraging GPU acceleration for unprecedented speed.

Training-free UAV navigation just got a major upgrade, achieving state-of-the-art results by leveraging structured spatial memory and a novel reasoning pipeline.

Many robotic policies that seem successful in manipulation tasks actually compromise safety, with SoftVTBench revealing a stark contrast between goal completion and physical safety metrics.

Each evaluation run in EVA-Client not only assesses performance but also enriches the training dataset, creating a continuous feedback loop for policy improvement.

TopoGPT achieves a remarkable leap in lane topology reasoning, producing geometrically consistent lane graphs that outperform existing methods by substantial margins.

A novel framework enables zero-shot localization across ground and drone views, outperforming traditional methods by leveraging a rich dataset and advanced geometric modeling.

MLLMs struggle with video temporal-logical reasoning, showing a substantial performance gap compared to human capabilities, especially as complexity increases.