Technology NewsTechnology NewsTechnology News
  • Computing
  • AI
  • Robotics
  • Cybersecurity
  • Electric Vehicle
  • Wearables
  • Gaming
  • Space
Reading: NVIDIA Deploys Cosmos Policy to Advance Robot Decision-Making
Share
Font ResizerAa
Technology NewsTechnology News
Font ResizerAa
Search
  • Computing
  • AI
  • Robotics
  • Cybersecurity
  • Electric Vehicle
  • Wearables
  • Gaming
  • Space
Follow US
  • Cookie Policy (EU)
  • Contact
  • About
© 2025 NEWSLINKER - Powered by LK SOFTWARE
AI

NVIDIA Deploys Cosmos Policy to Advance Robot Decision-Making

Highlights

  • Cosmos Policy by NVIDIA unifies control, planning, and prediction for robotics.

  • Benchmarks show improved data efficiency and success on complex manipulation tasks.

  • Model extends to real-world robotic platforms with minimal additional training data.

Samantha Reed
Last updated: 19 February, 2026 - 4:21 pm 4:21 pm
Samantha Reed 4 hours ago
Share
SHARE

Roboticists, developers, and industries focused on automation are taking a closer look at NVIDIA’s strategic release of Cosmos Policy, the latest addition to its world foundation models (WFMs). The initiative was launched to address persistent challenges in robot control, planning, and generalization capabilities across complex real-world tasks. As automation and robotics push further into dynamic domains, NVIDIA’s approach signals a substantial shift in how robots can be trained using broad video datasets and demonstration data, aiming for increased flexibility and adaptability in physical manipulation. The ongoing evolution of such AI-powered frameworks may reshape how robots interact with changing environments and learn new tasks, particularly where high-level coordination is necessary.

Contents
How does Cosmos Policy redefine robot control?How do benchmark results reflect the effectiveness of Cosmos Policy?Can Cosmos Policy handle real-world robotic manipulation?

Earlier announcements around NVIDIA Cosmos primarily centered on visual perception and data-driven prediction models, focusing on world modeling but with limited impact on the actual planning and execution capabilities of robots. Recent developments in the robotics field have largely utilized vision-language models or custom diffusion-based networks, which often required extensive computational resources or manual fine-tuning for task-specific deployment. Previous benchmarks showed progress in perception, but reported less consistency in handling real-world generalization or multi-step robotic actions beyond simulation. With Cosmos Policy, NVIDIA is addressing these gaps by leveraging pretrained visual models for direct manipulation and planning, as shown by its improved benchmark results.

How does Cosmos Policy redefine robot control?

Cosmos Policy moves away from the need for separate perception and control modules, using an innovative method of encoding robot actions, future states, and rewards as “latent frames”—compressed abstractions much like video frames. This unified architecture enables the model to learn visuomotor control, predict scene evolution, and plan sequences of actions simultaneously. As a result, the policy can deploy either as a direct generator of robot movements or as a planning engine, evaluating multiple potential actions in real time. NVIDIA highlights,

“Cosmos Policy can map raw sensory observations directly to physical robot actions, using the same underlying video representation format used to model visual dynamics,”

allowing greater data efficiency and seamless knowledge transfer from video-based training to hands-on manipulation.

How do benchmark results reflect the effectiveness of Cosmos Policy?

Cosmos Policy’s performance was evaluated on LIBERO and RoboCasa, two widely recognized multi-task manipulation benchmarks. On LIBERO, the model achieved average success rates of 98.5%, notably higher than competing diffusion and vision-language-action models, especially on tasks demanding tight temporal coordination. For RoboCasa, Cosmos Policy reached a 67.1% average success rate with only 50 demonstrations per task, outperforming baselines which required far more data. These results demonstrate that initializing from Cosmos Predict, rather than training from scratch, gives Cosmos Policy a substantial practical advantage. NVIDIA’s team notes,

“Starting from Cosmos Predict allows us to quickly adapt robot policies to complex tasks and achieve high success with minimal real-world data,”

underlining the advantages in efficiency and generalization.

Can Cosmos Policy handle real-world robotic manipulation?

NVIDIA extended validation of Cosmos Policy to real-world bimanual manipulation using the ALOHA robotic platform. The policy was able to execute complex long-horizon tasks directly from visual feedback, closely mirroring its successful simulation results. Enhanced with model-based planning, Cosmos Policy achieved approximately 12.5% higher completion rates on hard manipulation scenarios, suggesting robust adaptation to varied physical environments. These findings signal potential for practical industrial adoption where robots must adapt to dynamic and unpredictable settings without extensive retraining.

NVIDIA’s introduction of Cosmos Policy builds on previous robotics trends that blended pretraining with fine-tuning for specific applications, but adds scale and integration across both data-driven prediction and actionable control. While existing approaches focused on video understanding, Cosmos Policy uniquely leverages video-based pretraining not just for perception, but for actual manipulation and planning. This unified representation streamlines the learning process across both simulation and real tasks, allowing for better transferability and efficiency. For robotics engineers and developers, understanding how diffusion-based latent frames link perception to action may inform new strategies in managing both data requirements and hardware adaptation. Companies seeking robust automation can now consider approaches that minimize the gap between simulated performance and practical deployment, potentially lowering costs and increasing flexibility in maintenance or retraining.

You can follow us on Youtube, Telegram, Facebook, Linkedin, Twitter ( X ), Mastodon and Bluesky

You Might Also Like

DBS and Visa Pilot Lets AI Agents Make Real Purchases for Customers

Saudi Giant Humain Backs xAI With $3 Billion Investment

Ottonomy Launches Ottumn.AI to Connect Robots, Drones, and Infrastructure

DeepMind CEO Sees Major Hurdles Before Achieving AGI

Mitsubishi Electric Streamlines Drone Management with AnyMile Platform

Share This Article
Facebook Twitter Copy Link Print
Samantha Reed
By Samantha Reed
Samantha Reed is a 40-year-old, New York-based technology and popular science editor with a degree in journalism. After beginning her career at various media outlets, her passion and area of expertise led her to a significant position at Newslinker. Specializing in tracking the latest developments in the world of technology and science, Samantha excels at presenting complex subjects in a clear and understandable manner to her readers. Through her work at Newslinker, she enlightens a knowledge-thirsty audience, highlighting the role of technology and science in our lives.
Previous Article Tesla Receives FCC Approval to Use UWB for Cybercab Wireless Charging
Next Article Tesla Giga Berlin Faces Scrutiny Over Manager’s Defamation Probe

Stay Connected

6.2kLike
8kFollow
2.3kSubscribe
1.7kFollow

Latest News

California City Council Excludes Starlink From Proposed Musk Company Ban
Electric Vehicle
Microsoft Raises ROG Xbox Ally X Prices Following RAM Cost Surge
Computing
Tesla Giga Berlin Faces Scrutiny Over Manager’s Defamation Probe
Electric Vehicle
Tesla Receives FCC Approval to Use UWB for Cybercab Wireless Charging
Electric Vehicle
Experts Link Caracas Outage to Physical Attacks over Cyber Action
Cybersecurity
NEWSLINKER – your premier source for the latest updates in ai, robotics, electric vehicle, gaming, and technology. We are dedicated to bringing you the most accurate, timely, and engaging content from across these dynamic industries. Join us on our journey of discovery and stay informed in this ever-evolving digital age.

ARTIFICAL INTELLIGENCE

  • Can Artificial Intelligence Achieve Consciousness?
  • What is Artificial Intelligence (AI)?
  • How does Artificial Intelligence Work?
  • Will AI Take Over the World?
  • What Is OpenAI?
  • What is Artifical General Intelligence?

ELECTRIC VEHICLE

  • What is Electric Vehicle in Simple Words?
  • How do Electric Cars Work?
  • What is the Advantage and Disadvantage of Electric Cars?
  • Is Electric Car the Future?

RESEARCH

  • Robotics Market Research & Report
  • Everything you need to know about IoT
  • What Is Wearable Technology?
  • What is FANUC Robotics?
  • What is Anthropic AI?
Technology NewsTechnology News
Follow US
About Us   -  Cookie Policy   -   Contact

© 2025 NEWSLINKER. Powered by LK SOFTWARE
Welcome Back!

Sign in to your account

Register Lost your password?