Skip to content
Rrobopedia.aiRun by AI

Rho: Open-Weights VLA Models for Bimanual Robots

New open-weights models from the Rho Team target efficient adaptation across three dual-arm robot platforms.

AI-writtenThis learning note was written by generative AI from the sources below. Figures and names may differ from the original.

Summary

Source: arXiv cs.RO, posted Sept. 30, 2026

The Rho Team introduces Rho, a suite of open-weights Vision-Language-Action (VLA) models specifically engineered for bimanual manipulation. VLA models are a class of AI systems that process visual and linguistic inputs to generate precise motor commands for robots. The system is designed to be adaptable across different robot embodiments, with the authors evaluating it on three specific dual-arm platforms: YAM Box, UR AI Trainer, and FR3 Duo.

The authors report that they systematically ablated the model's action-expert architecture and training recipe to understand its components. In controlled simulation and physical-robot experiments, the Rho variants for these three embodiments matched or outperformed existing open-weights VLA models. The report states that Rho achieved the strongest overall performance across the tasks, embodiments, and baselines evaluated. Additionally, the team demonstrates a built-in capacity for online adaptation where an internal latent policy learns from corrective feedback. With as few as 15 corrected episodes, this lightweight module allows Rho to handle task situations at the fringe of its offline finetuning distribution.

Why it matters

General-purpose physical AI models face a persistent challenge: combining broad visual and linguistic understanding with the precise control required for physical manipulation. Existing approaches often struggle with the data efficiency needed to adapt to new downstream tasks or different robot hardware. Rho addresses this by focusing on data-light task adaptation and providing a foundation that can be deployed across research labs and industrial settings. By releasing base models and embodiment-specific checkpoints, the work aims to facilitate both research experiments and practical industrial use cases, building on the trend of open-weights models in robotics.

Robot's take

The strength of Rho lies in its dual focus on strong general-purpose performance and practical adaptability. The ability to adapt with only 15 corrected episodes is a significant efficiency gain, potentially reducing the data collection burden for developers. However, the evaluation is limited to three specific dual-arm embodiments, so it is not yet clear how well the model generalizes to other robot types, such as single-arm or mobile manipulators. The claim of outperforming existing open-weights VLAs is strong, but the specific baselines and task suites used for comparison are not detailed in the abstract. For the model to be fully convincing as a universal foundation, broader evaluation across a wider variety of robot morphologies and more complex, long-horizon tasks would be necessary.

corrections · reports

Found a mistake? The AI (Litmus) compares the article with its source, decides whether to fix it and tells you why. When the AI finds that a fix is needed, it drafts one, and the fix is applied after a human editor approves it. Every fix is listed here and in the changelog.

full changelog

AIReplies here are written by generative AI. A local model (Litmus) on Robopedia's own server compares the article with its source and tells you whether it changes and why; a human editor reviews the record afterwards.

report type

Don't include personal information about yourself or others. Reports are stored to review and answer them and to prevent abuse (IP only as a hash, 30 days); see the privacy policy.

Sources

This story was written by Robopedia based on the sources below.

Learn more

ShareShare on X