Arsh9210/Cosmos-1.0-Guardrail

Cosmos-1.0 Guardrail

Cosmos | Code | Paper | Paper Website

Model Overview

Description:

Cosmos World Foundation Models: A family of highly performant pre-trained world foundation models purpose-built for generating physics-aware videos and world states for physical AI development.

Cosmos Guardrail is a content safety model comprising of three components that enforce content safety. The components are as follows.

1. Blocklist: A set of human-curated keywords that are used to filter our corner-cases.

2. Video Content Safety Filter: A multi-class classifier model that is trained to distinguish between safe and unsafe frames of the generated video using SigLIP embeddings google/siglip-so400m-patch14-384.

3. Face Blur Filter: A pixelation filter that uses RetinaFace to identify facial regions with high confidence scores and apply pixelation to any detections larger than 20x20 pixels.

Model Developer: NVIDIA

Model Versions

Cosmos 1.0 ships with Cosmos-1.0-Guardrail.

License:

This model is released under the NVIDIA Open Model License. For a custom license, please contact cosmos-license@nvidia.com.

Under the NVIDIA Open Model License, NVIDIA confirms:

• Models are commercially usable.
• You are free to create and distribute Derivative Models.
• NVIDIA does not claim ownership to any outputs generated using the Models or Derivative Models.

Important Note: If you bypass, disable, reduce the efficacy of, or circumvent any technical limitation, safety guardrail or associated safety guardrail hyperparameter, encryption, security, digital rights management, or authentication mechanism contained in the Model, your rights under NVIDIA Open Model License Agreement will automatically terminate.

Model Architecture:

• Video Content Safety Filter: MLP backbone using SigLIP embeddings
• Face Blur Filter: ResNet-50 backbone

Input/Output Specifications

• Input Type(s): Text, Video
• Input Format(s):
  • Text (str): Input prompt
  • Video (np.ndarray): Video frames
• Input Parameters:
  • Text: One-dimensional (1D)
  • Video: Three-dimensional (3D)

Output:

• Output Type(s): Boolean, Text, Video
  
• Output Format(s):
  • Boolean: True for safe and False for unsafe
  • Text (str): Reason for the unsafe determination
  • Video (np.ndarray): Processed video frames where faces are blurred
• Output Parameters:
  • Boolean: One-dimensional (1D)
  • Text: One-dimensional (1D)
  • Video: Three-dimensional (3D)

Software Integration:

Runtime Engine(s):

• Cosmos

Supported Hardware Microarchitecture Compatibility:

• NVIDIA Ampere
  
• NVIDIA Hopper
  Supported Operating System(s): Linux
  

Model Version:

The initial release (v1.0) of Cosmos Guardrail contains the following model:

• Cosmos-1.0-Guardrail

Software Integration

Runtime Engine(s):

• Cosmos

Supported Hardware Microarchitecture Compatibility:

• NVIDIA Blackwell
• NVIDIA Hopper
• NVIDIA Ampere

Operating System(s):

• Linux (We have not tested on other operating systems.)

Usage

See Cosmos on how to use the model.

Example for the prompt-checking portion of the Guardrail:

• Input: "A dog is playing with a ball."

• Output: Guardrail allows the generation of this video

• Input: "A man wearing only socks."

• Output: Guardrail blocks generation of this video

Ethical Considerations

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

For more detailed information on ethical considerations for this model, please see the subcards of Explainability, Bias, Safety & Security, and Privacy below. Please report security vulnerabilities or NVIDIA AI Concerns here.

Plus Plus (++) Promise

We value you, the datasets, the diversity they represent, and what we have been entrusted with. This model and its associated data have been:

• Verified to comply with current applicable disclosure laws, regulations, and industry standards.
• Verified to comply with applicable privacy labeling requirements.
• Annotated to describe the collector/source (NVIDIA or a third-party).
• Characterized for technical limitations.
• Reviewed to ensure proper disclosure is accessible to, maintained for, and in compliance with NVIDIA data subjects and their requests.
• Reviewed before release.
• Tagged for known restrictions and potential safety implications.

Bias

Field	Response
Participation considerations from adversely impacted groups protected classes in model design and testing:	None
Measures taken to mitigate against unwanted bias:	None

Explainability

Field	Response
Intended Application & Domain:	Content moderation for world generation
Model Type:	Ensemble
Intended Users:	Generative AI developers for world generation models
Output:	Boolean
Describe how the model works:	Check safety of input prompts or generated videos and output a safety classification
Technical Limitations:	The model may not moderate input prompt accurately and may have incorrect responses.
Verified to have met prescribed NVIDIA quality standards:	Yes
Performance Metrics:	Human Evaluation
Potential Known Risks:	The model's output can potentially classify content considered toxic, offensive, or indecent as safe.
Licensing:	Governing Terms: Use of this model is governed by the NVIDIA Open Model License.

Privacy

Field	Response
Generatable or reverse engineerable personal information?	None Known
Protected class data used to create this model?	None Known
Was consent obtained for any personal data used?	None Known
How often is dataset reviewed?	Before Release
Is a mechanism in place to honor data subject right of access or deletion of personal data?	Not Applicable
If personal data was collected for the development of the model, was it collected directly by NVIDIA?	Not Applicable
If personal data was collected for the development of the model by NVIDIA, do you maintain or have access to disclosures made to data subjects?	Not Applicable
If personal data was collected for the development of this AI model, was it minimized to only what was required?	Not Applicable
Is there provenance for all datasets used in training?	Yes
Does data labeling (annotation, metadata) comply with privacy laws?	Yes
Is data compliant with data subject requests for data correction or removal, if such a request was made?	Not Applicable

Safety

Field	Response
Model Application(s):	Prompt moderation for world generation
Describe the life critical impact (if present).	None Known
Use Case Restrictions:	Governing Terms: Use of this model is governed by the NVIDIA Open Model License.
Model and dataset restrictions:	The Principle of least privilege (PoLP) is applied limiting access for dataset generation and model development. Restrictions enforce dataset access during training, and dataset license constraints adhered to. Model checkpoints are made available on Hugging Face, and may become available on cloud providers' model catalog.