7 papers
Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens
Ruiyang Si, Jianxin Bi, Shunyu Yang +12
Vision-language model (VLM) agents can control robots through visual feedback and action primitives, but repeated LLM calls and redundant observations incur substantial token overh…
Learning Physical Interaction: A Survey of Tactile- and Force-aware Robot Learning
Shilin Shan, Chuhao Zhou, Ruize Wang +30
Physically grounded robot intelligence requires robots to perceive, reason about, and regulate their interactions with the physical world. This capability is particularly critical…
Heterogeneous Tactile Transformer
Jianxin Bi, Qiang Wang, Jayaram Reddy +4
Tactile sensors are inherently heterogeneous: a model trained on one sensor cannot be directly used on another, which limits learning contact-rich manipulation policies from divers…
HumanScale: Egocentric Human Video Can Outperform Real-Robot Data for Embodied Pretraining
Juncheng Ma, Jianxin Bi, Yufan Deng +19
Embodied foundation models are expected to benefit from data scaling like large language models, but face a much tighter data bottleneck. Teleoperated real-robot trajectories remai…
VLA-Touch: Enhancing Vision-Language-Action Models with Dual-Level Tactile Feedback
Jianxin Bi, Kevin Yuchen Ma, Ce Hao +2
Tactile feedback is generally recognized to be crucial for effective interaction with the physical world. However, state-of-the-art Vision-Language-Action (VLA) models lack the abi…
Imitation Learning with Limited Actions via Diffusion Planners and Deep Koopman Controllers
Jianxin Bi, Kelvin Lim, Kaiqi Chen +2
Recent advances in diffusion-based robot policies have demonstrated significant potential in imitating multi-modal behaviors. However, these approaches typically require large quan…
Safety-Constrained Policy Transfer with Successor Features
Zeyu Feng, Bowen Zhang, Jianxin Bi +1
In this work, we focus on the problem of safe policy transfer in reinforcement learning: we seek to leverage existing policies when learning a new task with specified constraints.…