ABot-M0: VLA Foundation Model for Robotic Manipulation with Action Manifold Learning

2026-02-11

AI Summary

This work addresses the challenges of fragmented data, inconsistent representations, and misaligned training objectives faced by embodied agents operating across diverse hardware platforms. To this end, the authors propose Action Manifold Learning (AML), which leverages the action manifold hypothesis to project high-dimensional actions into a low-dimensional, continuous space constrained by physical priors, thereby enhancing policy stability and decoding efficiency. They introduce UniACT, a standardized dataset comprising 6 million trajectories and 9,500 hours of interaction data, and develop a dual-stream perception architecture that integrates semantic knowledge from vision-language models with 3D geometric priors, enabling end-to-end policy learning via a DiT backbone. Experiments demonstrate that each component is independently effective and collectively yields significant, additive performance gains, markedly improving cross-platform and cross-task generalization across six public benchmarks.

Technology Category

• Multiagent Systems: Multiagent Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Intelligent Robots: Embodied AI

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Building general-purpose embodied agents across diverse hardware remains a central challenge in robotics, often framed as the''one-brain, many-forms''paradigm. Progress is hindered by fragmented data, inconsistent representations, and misaligned training objectives. We present ABot-M0, a framework that builds a systematic data curation pipeline while jointly optimizing model architecture and training strategies, enabling end-to-end transformation of heterogeneous raw data into unified, efficient representations. From six public datasets, we clean, standardize, and balance samples to construct UniACT-dataset, a large-scale dataset with over 6 million trajectories and 9,500 hours of data, covering diverse robot morphologies and task scenarios. Unified pre-training improves knowledge transfer and generalization across platforms and tasks, supporting general-purpose embodied intelligence. To improve action prediction efficiency and stability, we propose the Action Manifold Hypothesis: effective robot actions lie not in the full high-dimensional space but on a low-dimensional, smooth manifold governed by physical laws and task constraints. Based on this, we introduce Action Manifold Learning (AML), which uses a DiT backbone to predict clean, continuous action sequences directly. This shifts learning from denoising to projection onto feasible manifolds, improving decoding speed and policy stability. ABot-M0 supports modular perception via a dual-stream mechanism that integrates VLM semantics with geometric priors and multi-view inputs from plug-and-play 3D modules such as VGGT and Qwen-Image-Edit, enhancing spatial understanding without modifying the backbone and mitigating standard VLM limitations in 3D reasoning. Experiments show components operate independently with additive benefits. We will release all code and pipelines for reproducibility and future research.

Problem

• embodied agents
• robotic manipulation
• general-purpose intelligence
• action prediction
• cross-platform generalization

Innovation

• Action Manifold Learning
• VLA Foundation Model
• Unified Robotic Dataset
• DiT-based Action Prediction
• Modular 3D Perception