FLOWER: Democratizing Generalist Robot Policies with Efficient Vision-Language-Action Flow Policies

2025-09-05

AI Summary

Existing diffusion-based vision-language-action (VLA) policies rely on multi-billion-parameter architectures and massive datasets, incurring prohibitive computational costs that hinder real-world robot deployment. To address this, we propose an efficient VLA model featuring an intermediate modality fusion mechanism and action-specific Global-AdaLN conditioning modules—enabling significant parameter reduction. We further integrate LLM layer pruning with a modular diffusion architecture, enabling full pretraining within only 200 H100 GPU-hours. Our model contains just 950 million parameters yet achieves a state-of-the-art 4.53 success rate on the CALVIN ABC benchmark. Moreover, it matches or exceeds the performance of substantially larger models across 190 diverse tasks spanning both simulation and real-world robotic platforms. This work marks the first demonstration of a lightweight, high-performance, and strongly generalizable universal robot controller.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Developing efficient Vision-Language-Action (VLA) policies is crucial for practical robotics deployment, yet current approaches face prohibitive computational costs and resource requirements. Existing diffusion-based VLA policies require multi-billion-parameter models and massive datasets to achieve strong performance. We tackle this efficiency challenge with two contributions: intermediate-modality fusion, which reallocates capacity to the diffusion head by pruning up to $50%$ of LLM layers, and action-specific Global-AdaLN conditioning, which cuts parameters by $20%$ through modular adaptation. We integrate these advances into a novel 950 M-parameter VLA called FLOWER. Pretrained in just 200 H100 GPU hours, FLOWER delivers competitive performance with bigger VLAs across $190$ tasks spanning ten simulation and real-world benchmarks and demonstrates robustness across diverse robotic embodiments. In addition, FLOWER achieves a new SoTA of 4.53 on the CALVIN ABC benchmark. Demos, code and pretrained weights are available at https://intuitive-robots.github.io/flower_vla/.

Problem

• Developing efficient vision-language-action policies for robotics
• Reducing computational costs and resource requirements
• Achieving competitive performance with smaller models

Innovation

• Prunes 50% LLM layers for efficiency
• Uses Global-AdaLN conditioning to cut parameters
• Integrates advances into 950M-parameter VLA model