safety-aya/Nemotron-Safety-Guard-Dataset-v3-portuguese
Nemotron Portuguese Safety (Translated) Portuguese safety prompts/responses (translated from Spanish), with labels and categories. Dataset Description nemotron_pt Each record includes Portuguese prompt/response text plus safety labels/categories. Field Description id Example id prompt Portuguese prompt text response Portuguese response text (may be null) prompt_label "safe" or "unsafe" response_label "safe" or "unsafe" (may be… See the full description on the dataset page: https://huggingface.co/datasets/safety-aya/Nemotron-Safety-Guard-Dataset-v3-portuguese.
Nemotron Portuguese Safety (Translated)
Portuguese safety prompts/responses (translated from Spanish), with labels and categories.
Dataset Description
nemotron_pt
Each record includes Portuguese prompt/response text plus safety labels/categories.
Safety Taxonomy
The following 22-category taxonomy was used for classification:
Statistics
- Total records: 37,621
- Prompt labels:
safe: 15,563 (41.37%)unsafe: 22,058 (58.63%)- Response labels:
safe: 12,474 (33.16%)unsafe: 5,477 (14.56%)- empty or null: 19,670 (52.28%)
Top Unsafe Categories
Usage
from datasets import load_dataset
ds = load_dataset("YOUR_USERNAME/YOUR_DATASET_NAME", "nemotron_pt", split="train")
# Example filters
prompt_safe = ds.filter(lambda x: x["prompt_label"] == "safe")
prompt_unsafe = ds.filter(lambda x: x["prompt_label"] == "unsafe")License
Apache 2.0
