# **Explainability**

|Field:|Response:|  
|:---:|:---:|  
|Intended Domain:| Open healthcare foundation model for generalized surgical and ultrasound robotics reasoning and skills.|  
|Model Type: |Robot VLA|  
|Intended Users:|Researchers and developers working on surgical robotics and ultrasound applications.|  
|Output:|The model outputs are actions, and the units are floating-point values. This is referred to as "robot action policy." Actions consist of continuous-value vectors that correspond to different motor controls on a robot.|
|Describe how the model works:|Accepts multimodal inputs such as video, ultrasound, proprioception, and language, then outputs a robot action policy.|  
|Technical Limitations & Mitigation:| This model is not tested or intended for use in mission critical or clinical applications that require functional safety. Use in those applications is at the user's own risk and sole responsibility, including taking the necessary steps to add needed guardrails or safety mechanisms prior to deployment. More generally, limitations include, but are not limited to:<br>- The model may underperform in operating room environments or device configurations that differ from the training distribution.<br>- Coverage may be limited for rare procedures, uncommon instruments, specialized workflows, or underrepresented institutions.<br>- Performance can vary across sites, sensors, and embodiment interfaces; additional fine-tuning and validation may be required for new deployments.<br><br>Risks and possible mitigations include:<br>Risk: Model underperformance in variable operating room conditions, device configurations, or imaging settings.<br>Mitigation: Expand data coverage across devices/settings and fine-tune for target environments.<br><br>Risk: Integration challenges across surgical/ultrasound platforms with different control interfaces or sensing configurations.<br>Mitigation: Provide embodiment-specific integration guidance and validation procedures.<br><br>Risk: Limited coverage for rare procedures or uncommon instrument/tooling setups.<br>Mitigation: Curate targeted data and evaluate on representative task subsets.|  
|Verified to have met prescribed quality standards?|Yes|  
|Performance Metrics:|Success rate, as well as the following:<br>1) if the trajectory is smooth and does not jitter<br>2) if the robot does not hit any other objects<br>3) if the trajectory is natural|  
|Potential Known Risks:|This model is not tested or intended for clinical applications that require functional safety. The use of the model in those applications is at the user's own risk and sole responsibility, including taking the necessary steps to add needed guardrails or safety mechanisms prior to deployment.|  
|End User License Agreement:| Your use of this model is governed by the [NSCL V1 License](https://developer.download.nvidia.com/licenses/NVIDIA-OneWay-Noncommercial-License-22Mar2022.pdf).|