Zhengjie Zhang

A deterministic 25 Hz tactile reflex controller is used as a collection-time teacher that shapes grasping behavior in demonstrations for policy learning, while disturbance rejection remains controller-dependent, revealing the boundary of tactile-free policy.

Riemannian Geometry-Sensitive Quantization (RGSQ), which formulates quantization as a reconstruction problem under a unified Fisher-Riemannian metric, enabling standard unimodal PTQ methods to evaluate multimodal quantization error under their original assumptions.

Bridging the gap between simulation and real-world performance, DeepInsight II reveals how standardized evaluation can enhance the reliability of embodied AI systems.

Achieve state-of-the-art colo-segment recognition by combining edge-guided spatial feature extraction with a novel temporal attention mechanism, outperforming existing methods by a large margin.