AtomVLA: Scalable Post-Training for Robotic Manipulation via Predictive Latent World Models

2026-03-09

AI Summary

This work addresses the challenge of error accumulation in long-horizon tasks faced by existing vision-language-action (VLA) models due to the absence of intermediate guidance. The authors propose the first subtask-aware VLA framework, which leverages a large language model to decompose high-level instructions into atomic subtasks and employs a pretrained predictive latent world model to evaluate the alignment between action segments and subtask goals in latent space. By integrating offline post-training with Group Relative Policy Optimization, the method effectively suppresses error propagation without requiring costly online interaction. The approach achieves state-of-the-art average success rates of 97.0% on LIBERO and 48.0% on the more challenging LIBERO-PRO benchmark, and demonstrates strong generalization to complex long-horizon tasks on the Galaxea R1 Lite real-world robot platform.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Computer Vision: Language and Vision
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Vision-Language-Action (VLA) models demonstrate remarkable potential for generalizable robotic manipulation. The execution of complex multi-step behaviors in VLA models can be improved by robust instruction grounding, a critical component for effective control. However, current paradigms predominantly rely on coarse, high-level task instructions during supervised fine-tuning. This instruction grounding gap leaves models without explicit intermediate guidance, leading to severe compounding errors in long-horizon tasks. Therefore, bridging this instruction gap and providing scalable post-training for VLA models is urgent. To tackle this problem, we propose \method, the first subtask-aware VLA framework integrated with a scalable offline post-training pipeline. Our framework leverages a large language model to decompose high-level demonstrations into fine-grained atomic subtasks. This approach utilizes a pretrained predictive world model to score candidate action chunks against subtask goals in the latent space, mitigating error accumulation while significantly improving long-horizon robustness. Furthermore, this approach enables highly efficient Group Relative Policy Optimization without the prohibitive expenses associated with online rollouts on physical robots. Extensive simulations validate that our AtomVLA maintains strong robustness under perturbations. When evaluated against fundamental baseline models, it achieves an average success rate of 97.0\% on the LIBERO benchmark and 48.0\% on the LIBERO-PRO benchmark. Finally, experiments conducted in the real world using the Galaxea R1 Lite platform confirm its broad applicability across diverse tasks, especially long-horizon tasks. All datasets, checkpoints, and code will be released to the public domain following the acceptance of this work for future research.

Problem

• Vision-Language-Action
• instruction grounding
• long-horizon tasks
• error compounding
• robotic manipulation

Innovation

• Vision-Language-Action (VLA)
• predictive latent world model
• subtask decomposition
• offline post-training
• long-horizon robotic manipulation