By using the Model(s), you agree to the Model License(s).
Use the NGC CLI to download:
Copied!
Model
NVIDIA Nemotron-Parse v2.0 - document understanding VLM: image in, structured text (with bboxes and semantic classes) out.
By using the Model(s), you agree to the Model License(s).
Use the NGC CLI to download:
Copied!
Bias
| Field | Response |
|---|---|
| Participation considerations from adversely impacted groups protected classes in model design and testing: | None |
| Measures taken to mitigate against unwanted bias: | Dataset curation, filtering, and evaluation include review across document types, layouts, languages, and scripts where available. Users should evaluate performance on their target document domains before deployment. |
| Bias Metric (If Measured): | Not measured as a demographic fairness metric. Model quality is evaluated through document parsing, OCR, layout, table, chart, and reading-order metrics. |
| (For GPAI Models) Which characteristic (feature) show(s) the greatest difference in performance?: | Pending final release evaluation. Expected variation should be assessed across language, script, image quality, document domain, table complexity, and chart type. |
| (For GPAI Models): Which feature(s) have have the worst performance overall? | Pending final release evaluation. Known risk areas include low-resolution scans, handwriting, dense tables, chart-heavy pages, rare scripts, decorative fonts, and unusual layouts. |
| (For GPAI Models): If using internal data, description of methods implemented in data acquisition or processing, if any, to address the prevalence of identifiable biases in the training, testing, and validation data: | Training and evaluation data are curated to cover diverse document layouts, document domains, languages, scripts, and synthetic/rendered document conditions where available. |
| (For GPAI Models): Tools used to assess statistical imbalances and highlight patterns that may introduce bias into AI models: | Internal dataset audits, benchmark stratification, OCR/layout evaluation, ParseBench, IndicVisionBench, MOSCAR, and qualitative review. |
| (For GPAI Models): Tools used to assess statistical imbalances and highlight patterns that may introduce bias into AI models: | The datasets used for document parsing evaluation do not collectively or exhaustively represent all document domains, languages, scripts, layouts, or image-quality conditions. To mitigate this, users should evaluate NVIDIA Nemotron Parse 2.0 on representative target documents and consider additional evaluation, fine-tuning, filtering, or human review for their intended workflow. |