Pure Vision Language Action (VLA) Models: A Comprehensive Survey

2025-09-23

AI Summary

Existing vision-language-action (VLA) models lack a unified conceptual framework, hindering systematic understanding of how VLA enables robots to transition from passive perception to active decision-making and physical interaction. Method: We propose a multidimensional taxonomy—spanning autoregressive, diffusion-based, reinforcement learning, and hybrid paradigms—and conduct an integrative analysis of 300+ state-of-the-art studies to characterize modeling motivations, technical pathways, and applicability boundaries for each paradigm. Contribution/Results: This work delivers the first comprehensive VLA survey, encompassing benchmark tasks, simulation platforms, and open-source datasets. It identifies core challenges—including scalability, embodied generalization, and world modeling—and charts an evolutionary trajectory toward general-purpose embodied agents. By synthesizing theoretical foundations and empirical advances, our survey provides a structured knowledge graph and methodological foundation for both VLA research and embodied intelligence development.

Technology Category

• Intelligent Robots: Embodied AI
• Computer Vision: Language and Vision
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

The emergence of Vision Language Action (VLA) models marks a paradigm shift from traditional policy-based control to generalized robotics, reframing Vision Language Models (VLMs) from passive sequence generators into active agents for manipulation and decision-making in complex, dynamic environments. This survey delves into advanced VLA methods, aiming to provide a clear taxonomy and a systematic, comprehensive review of existing research. It presents a comprehensive analysis of VLA applications across different scenarios and classifies VLA approaches into several paradigms: autoregression-based, diffusion-based, reinforcement-based, hybrid, and specialized methods; while examining their motivations, core strategies, and implementations in detail. In addition, foundational datasets, benchmarks, and simulation platforms are introduced. Building on the current VLA landscape, the review further proposes perspectives on key challenges and future directions to advance research in VLA models and generalizable robotics. By synthesizing insights from over three hundred recent studies, this survey maps the contours of this rapidly evolving field and highlights the opportunities and challenges that will shape the development of scalable, general-purpose VLA methods.

Problem

• Surveying Vision Language Action models for robotics control in dynamic environments
• Classifying VLA approaches into different paradigms and analyzing their implementations
• Identifying key challenges and future directions for generalizable robotics research

Innovation

• VLA models transform passive VLMs into active agents
• Classifies VLA approaches into autoregression, diffusion, and hybrid methods
• Analyzes applications, datasets, and future challenges for generalizable robotics