Do World Action Models Generalize Better than VLAs? A Robustness Study

2026-03-23

AI Summary

This work presents the first systematic comparison of World Action Models (WAMs) and Vision–Language–Action models (VLAs) in terms of robustness and generalization across unseen environments, evaluating them on the LIBERO-Plus and RoboTwin 2.0-Plus benchmarks. Using large-scale video-pretrained WAMs—such as Cosmos-Policy and LingBot-VA—and state-of-the-art VLA policies like π₀.₅, the study conducts end-to-end assessments under visual–language perturbations. Results demonstrate that WAMs substantially outperform most data-hungry VLAs, with LingBot-VA achieving a 74.2% success rate on RoboTwin 2.0-Plus and Cosmos-Policy reaching 82.2% on LIBERO-Plus. These findings underscore the critical role of video-pretrained dynamic priors in enhancing generalization for real-world robotic manipulation tasks.

Technology Category

• Intelligent Robots: Manipulation
• Computer Vision: Language and Vision
• Natural Language Processing: Safety and Robustness

Application Category

• 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
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Robot action planning in the real world is challenging as it requires not only understanding the current state of the environment but also predicting how it will evolve in response to actions. Vision-language-action (VLA), which repurpose large-scale vision-language models for robot action generation using action experts, have achieved notable success across a variety of robotic tasks. Nevertheless, their performance remains constrained by the scope of their training data, exhibiting limited generalization to unseen scenarios and vulnerability to diverse contextual perturbations. More recently, world models have been revisited as an alternative to VLAs. These models, referred to as world action models (WAMs), are built upon world models that are trained on large corpora of video data to predict future states. With minor adaptations, their latent representation can be decoded into robot actions. It has been suggested that their explicit dynamic prediction capacity, combined with spatiotemporal priors acquired from web-scale video pretraining, enables WAMs to generalize more effectively than VLAs. In this paper, we conduct a comparative study of prominent state-of-the-art VLA policies and recently released WAMs. We evaluate their performance on the LIBERO-Plus and RoboTwin 2.0-Plus benchmarks under various visual and language perturbations. Our results show that WAMs achieve strong robustness, with LingBot-VA reaching 74.2% success rate on RoboTwin 2.0-Plus and Cosmos-Policy achieving 82.2% on LIBERO-Plus. While VLAs such as $π_{0.5}$ can achieve comparable robustness on certain tasks, they typically require extensive training with diverse robotic datasets and varied learning objectives. Hybrid approaches that partially incorporate video-based dynamic learning exhibit intermediate robustness, highlighting the importance of how video priors are integrated.

Problem

• robot action planning
• generalization
• robustness
• vision-language-action models
• world action models

Innovation

• World Action Models
• Vision-Language-Action Models
• Robustness Evaluation
• Video Pretraining
• Dynamic Prediction