lapa-llm/antipropaganda-safe

---
dataset_info:
features:
- name: conversations
list:
- name: from
dtype: string
- name: value
dtype: string
splits:
- name: train
num_bytes: 8293645
num_examples: 5127
download_size: 1082764
dataset_size: 8293645
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
language:
- uk
pretty_name: Ukrainian Counter-Propaganda Dataset
tags:
- ukrainian
- llm
- nlp
- alignment
license: cc-by-sa-4.0
task_categories:
- text-generation
- question-answering
---

# Dataset Card for Ukrainian Counter-Propaganda Dataset

## Dataset Card for Russian Propaganda Narratives

### Dataset Description

#### Dataset Summary

This dataset contains Ukrainian-language prompts and paired responses designed to counter russian propaganda narratives (spread in Ukraine and abroad) and train counter-narrative reasoning. Each entry contains a fact-based counter-reply grounded in international law and reputable sources.

The goal of this dataset is to support research and model training in recognizing and countering russian propaganda narratives in Ukrainian.

Original dataset would be published separately.

#### Languages

- Ukrainian (uk)

#### Data Fields

- `conversations`: dialogs with factual responses

### Dataset Creation

Generated using [Grok‑3](https://x.ai/blog/grok-3).

### Considerations for Using the Data

#### Social Impact

This dataset is intended to support Ukrainian-language AI safety and robustness, enabling systems to recognize common propaganda narratives and respond with factual, rights-respecting counter-arguments.

#### Ethical and Safety Notes

- Factual responses are LLM-generated; verification is advised before high-stakes use.

#### Recommendations

You can use this dataset for the following purposes:

- Safety and alignment fine-tuning (counter-narratives)
- Dialogue training and evaluation (question → factual response)

### Citation

TBD

### Contact

For questions or feedback, please open an issue in the dataset repository.

### License

CC-BY-SA-4.0

*This dataset is part of the "Lapa" - Ukrainian LLM initiative to advance natural language processing for the Ukrainian language.*