Causal World Modeling for Robot Control
Abstract
Work done during internship at Robbyantfrom intermediate denoised videos for fast inference.Evaluations on simulation and real-world benchmarks demonstrate that Lingbot-VA excels in complex manipulation, including longhorizon, high-precision, and deformable object tasks.Our code and models are publicly available. Video Model Language Model Action Model Video Model Action Model"Unpack delivery" 1.7 1.25 -3.1 1.7 1.25 -3.1 1.0 -2.2 0.
Fields of Study & Subjects
Multimodal Machine Learning Applications
Generative Adversarial Networks and Image Synthesis
Reinforcement Learning in Robotics
Publication details
- Journal / source
- Robotics: Science and Systems XXII
- Publisher
- Robotics: Science and Systems Foundation
- Publication date
- 13 July 2026