| Field | Response |
| :---------------------------------------------------|:---------------------------------- |
| Model Application Field(s): | Customer Service, Media & Entertainment, Healthcare (Document Intelligence), Financial Services (Document Intelligence)<br> | |
| Describe the life critical impact (if present) | Not Applicable |
| Description of methods implemented in data acquisition or processing, if any, to address other types of potentially harmful data in the training, testing, and validation data: | Training data undergoes multi-stage processing including deduplication, filtering, and quality assessment. Data integrity filters are applied to chat, social sciences, and safety domains. Safety evaluations were conducted using AEGIS v2, Garak, RTVLM, and VLGuard, as well as NVIDIA Custom Datasets focused on child safety and dehumanization risks. RL recipes for security improvements are applied during post-training. Ablation studies for safety were conducted, and video safety and omni-specific benchmarks are investigated. An over-refusal evaluation is also performed to balance safety with model utility. |
| Description of any methods implemented in data acquisition or processing, if any, to address illegal or harmful content in the training data, including, but not limited to, child sexual abuse material (CSAM) and non-consensual intimate imagery (NCII) | Content safety is evaluated using NVIDIA Custom Datasets focused on child safety and dehumanization risks. AEGIS 1.0 evaluation covers 13 content safety categories with a required >=95% overall score AND 0 items under the sexual (minor) category. Qualitative review is conducted for every prompt/response under the suicide and self-harm category. Safety evaluations are run with both Reasoning ON and OFF modes. |
| Use Case Restrictions: | Governing Terms: Use of this model is governed by the [NVIDIA Open Model Agreement](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-agreement/)<br>| |
| 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. <br> | |
| Responsible data handling: | This AI model was developed based on our policies to ensure responsible data handling and risk mitigation. The datasets used for training have been scanned for harmful content and illegal content, consistent with our policies including scanning for Child Sexual Abuse Material (CSAM). Ongoing review and monitoring mechanisms are in place based on our policies and to maintain data integrity. |
|
|
|
|
