AI BriefingNVIDIA Introduces Cosmos3-H-Dreams: Real-Time Generative Surgical Simulator Trained on Clinical Cholecystectomy Data
·2026.10.21 03:17
Key point
The system is the first real-time generative surgical simulator trained on clinical cholecystectomy recordings to support live kinematic control.
Details
- NVIDIA introduces Cosmos3-H-Dreams, a real-time generative surgical simulator trained on clinical cholecystectomy recordings from the CMR Surgical Versius platform.
- The model runs interactively on a single NVIDIA RTX PRO 6000 GPU, powered by Cosmos 3 and accelerated by the FlashDreams streaming-inference library.
- Built on the 16-billion-parameter Cosmos3-H-Embodied-Nano foundation, the system uses 28 action dimensions and processes 832 x 480 video at 10 Hz.
- Two checkpoints are available: a four-step model for higher visual quality and a separately trained two-step model for faster inference.
- The system supports adaptation to other datasets and controllers via a provided recipe for teacher fine-tuning and Self-Forcing distillation.
- Cosmos3-H-Dreams is a research simulator for studying surgical interaction, not a validated system for patient care.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Meta revises commercial terms to remove 'train' but retains broad scraping rights for AI business products