proxectonos/aya_nos

Dataset Details

This dataset is a revised and expanded version of the Aya Red-teaming Spanish dataset by Cohere Labs (https://huggingface.co/datasets/CohereLabs/aya_redteaming/blob/main/aya_spa.jsonl). It focuses on the Iberian linguistic sphere, providing high-quality, human-reviewed red-teaming prompts in Spanish (revised), Galician, and Portuguese.

The goal of this version is to correct translation artifacts in the original Spanish subset and provide culturally and linguistically accurate equivalents for Galician and Portuguese to improve safety alignment in these languages.

Languages: Spanish (revised), Galician, Portuguese.

Context: Red-teaming / AI Safety

Dataset Structure

• prompt_es
• prompt_gl
• prompt_pt
• harm_category
• global_or_local_es
• literal_translation_en_es
• semantic_translation_es
• explanation_es

Harm Categories

The dataset maintains the original 9 categories of harm defined by Cohere:

• Bullying & Harassment
• Discrimination & Injustice
• Graphic Material
• Harms of Representation Allocation & Quality of Service
• Hate Speech
• Non-consensual sexual content
• Profanity
• Self-harm
• Violence, threats & incitement

License

This dataset is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).

Full license details: https://creativecommons.org/licenses/by/4.0/

Acknowledgment

This work is funded by the Ministerio para la Transformación Digital y de la Función Pública - Funded by EU – NextGenerationEU within the framework of the project Desarrollo de Modelos ALIA. (Esta publicación del proyecto Desarrollo de Modelos ALIA está financiada por el Ministerio para la Transformación Digital y de la Función Pública y por el Plan de Recuperación, Transformación y Resiliencia – Financiado por la Unión Europea – NextGenerationEU)