1. From Surgical World Model to Interactive Simulator 2. Distilling Cosmos-H-Surgical-Simulator for Real Time 2.1. A Surgical Teacher 2.2. Causal Warmup 2.3. Self-Forcing Distillation 3. FlashDreams: The Real-Time Inference Engine 4. Adapting to Your Own Data 5. What Is Next: Toward Closed-Loop Surgical Physical AI 6. Get Started Today Surgical robotics is moving quickly from teleoperation toward increasingly capable vision-language-action policies. But evaluating and training these systems remains difficult. Physical robotic platforms are expensive to operate, experiments are slow to reproduce, and failures can damage instruments or biological material. Conventional simulators provide a safer alternative, but surgical scenes are exceptionally difficult to model: deformable tissue, fine instrument interactions, specular surfaces, sutures, needles, smoke, and occlusions all matter.
NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics
A Blog post by NVIDIA on Hugging Face
HF Hugging Face Blog 
Key points
- Today, we are introducing the next step: Cosmos-H-Dreams, a real-time, action-conditioned generative simulator for surgical robotics.
- Cosmos-H-Dreams distills the capabilities of Cosmos-H-Surgical-Simulator into a causal, few-step student model and serves it through FlashDreams, NVIDIA's accelerated streaming-inference library.
- Cosmos-H-Surgical-Simulator is an action-conditioned world foundation model built on NVIDIA Cosmos-Predict2.5-2B and post-trained on the Open-H-Embodiment dataset.
Sentences selected automatically from the original article by Hugging Face Blog.
Read the full story on Hugging Face Blog That's the opening of the story. The full piece is published by Hugging Face Blog.
Continue reading ↗ Story details
- Published
- Format
- Article
- Original
- huggingface.co ↗
Related reporting
2 stories on this topic from 2 sources, oldest first.
- Hugging Face Blog NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics (this story)
