Institution profile

AgiBot

Industry researchasia · cn
Official website
Research library41linked papers
Opportunities0open roles
Selected work

Representative Papers

SOP: A Scalable Online Post-Training System for Vision-Language-Action Models

Jan 06, 2026arXiv.org

This work proposes SOP, the first online, multi-robot collaborative, and multi-task post-training framework for general-purpose vision-language-action (VLA) models. Existing VLA post-training methods are typically offline, single-machine, or task-specific, limiting their capacity for efficient online adaptation and large-scale real-world learning. SOP addresses this gap through a closed-loop bitstream architecture that tightly couples a fleet of robots with a cloud-based learner. The system integrates interactive imitation learning (HG-DAgger) and reinforcement learning (RECAP), enabling asynchronous policy updates and human-in-the-loop interventions. Evaluated on real-world tasks such as cloth folding and box assembly, SOP significantly improves pretrained model performance within hours, with gains scaling nearly linearly with the number of robots while preserving the generality of a single shared policy.

4 citationsRead paper

ACoT-VLA: Action Chain-of-Thought for Vision-Language-Action Models

Jan 16, 2026

This work addresses the limitation of existing vision-language-action models in preserving fine-grained information during intermediate reasoning, which constrains policy accuracy. To overcome this, the authors propose the Action Chain-of-Thought (ACoT) paradigm, which explicitly models intermediate reasoning as a sequence of coarse-grained action intentions. ACoT integrates an Explicit Action Reasoner (EAR) that generates reference trajectories and an Implicit Action Reasoner (IAR) that extracts latent action priors from multimodal inputs, enabling structured reasoning in the action space. Evaluated on LIBERO, LIBERO-Plus, and VLABench, the method achieves success rates of 98.5%, 84.1%, and 47.4%, respectively, substantially outperforming current state-of-the-art approaches.

1 citationsRead paper
Recent publications

Latest Papers