subcards

physical_plausibility_evaluator_subcards/bias.md

@@ -0,0 +1,9 @@
+# Bias
+
+| Field | Response |
+| --- | --- |
+| Participation considerations from adversely impacted groups [protected classes][protected] in model design and testing: | Not Applicable |
+| Measures taken to mitigate against unwanted bias: | Not Applicable |
+| Bias Metric (If Measured): | Not measured. |
+
+[protected]: https://calcivilrights.ca.gov/disputeresolution/protected-characteristics/

physical_plausibility_evaluator_subcards/explainability.md

@@ -0,0 +1,15 @@
+# Explainability
+
+| Field | Response |
+| --- | --- |
+| Intended Task/Domain: | Video quality screening using task-specific good/bad labels. |
+| Model Type: | Frozen video transformer, learned attention pooling and binary linear classifier. |
+| Intended Users: | Developers and researchers evaluating or filtering video datasets. |
+| Output: | Bad-class logit and sigmoid scores and the thresholded good-or-bad decision. |
+| Describe how the model works: | Frozen V-JEPA 2.1 encodes video into space-time tokens. A learned query pools normalized tokens into one vector; a linear classifier produces a bad logit. A calibrated threshold determines the decision. |
+| List the groups (or group characteristics) for which this was tested to produce comparable outcomes: | Not Applicable |
+| Technical Limitations & Mitigation: | Performance can change with scenes, generators, cameras, duration and labels. Sparse sampling can miss brief defects. Evaluate representative held-out videos, separate related scenes, and recalibrate on target data. Attention weights are not validated explanations or localization masks. |
+| Verified to have met prescribed NVIDIA quality standards: | Yes |
+| Performance Metrics: | Bad-class ROC AUC, bad recall and good-video rejection. |
+| Potential Known Risks: | Missed defects, rejection of useful videos and misleading confidence under distribution shift. Scores do not certify physical correctness. Class balancing does not establish demographic fairness or comparable outcomes across groups. |
+| Licensing: | [OpenMDW-1.1](https://openmdw.ai/license/1-1/) |

physical_plausibility_evaluator_subcards/privacy.md

@@ -0,0 +1,20 @@
+# Privacy
+
+| Field | Response |
+| :--- | :--- |
+| Generatable or reverse engineerable personal data*? | No |
+| Personal data used to create this model? | No |
+| Was consent obtained for any personal data* used? | Not Applicable |
+| How often is dataset reviewed? | During dataset preparation and annotation. |
+| 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 |
+| Was data from user interactions with the AI model (e.g. user input and prompts) used to train the model? | No |
+| Is there provenance for all datasets used in training? | Yes |
+| Does data labeling (annotation, metadata) comply with privacy laws? | Not Applicable |
+| Is data compliant with data subject requests for data correction or removal, if such a request was made? | Not Applicable |
+| Applicable Privacy Policy | [NVIDIA Privacy Policy](https://www.nvidia.com/en-us/about-nvidia/privacy-policy/) |
+
+*Responses cover the classifier dataset: selected FineVideo source frames/captions, generated videos and annotations.
+They do not characterize Meta's separate encoder-pretraining datasets.

physical_plausibility_evaluator_subcards/safety_security.md

@@ -0,0 +1,8 @@
+# Safety and Security
+
+| Field | Response |
+| --- | --- |
+| Model Application Field(s): | Video-level quality assessment and filtering using task-specific acceptability criteria. |
+| Describe the life critical impact: | Not intended as a safety-critical decision system or certification of physical correctness. |
+| Use Case Restrictions: | Abide by [OpenMDW-1.1](https://openmdw.ai/license/1-1/). |
+| Model and dataset restrictions: | Access to the model and dataset during generation and development was restricted to people who needed it, following the principle of least privilege. The dataset will not be released. |