Search papers, labs, and topics across Lattice.
School of Electrical, This work was supported by NTUitive Gap Fund (NGF-2025-17006) and National Research Foundation, Singapore, under its Medium-Sized Center for Advanced Robotics Technology Innovation. Zhongyuang Liu is with CertaintyX. Min He, Shaonan Yu, Xinhang Xu, Jianping Li and Lihua Xie are with the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798. Jianfei Yang is with the School of Mechanical and Aerospace Engineering, and jointly with the School of Electrical and Electronic Engineering at Nanyang Technological University, Singapore. Muqing Cao is with the Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213 USA. (Zhongyuang Liu and Min He contributed equally to this work. Jianping Li is the corresponding author. E-mail: jianping.li@ntu.edu.sg)
CMU Machine Learning4
0
5
10
LLMs can navigate complex 3D environments more effectively and with far fewer tokens by using a hierarchical scene graph representation derived from omnidirectional sensor data.
LLMs in embodied environments get a massive boost from structured rules, with rule retrieval alone contributing +14.9 pp to single-trial success.
Panoramic depth perception and differentiable physics unlock surprisingly robust collision avoidance, even generalizing to unseen simulation environments.
Legged robots can now tiptoe around your expensive gadgets, thanks to a new RL framework that combines semantic understanding with low-level control to avoid stepping on designated objects.
