--- license: apache-2.0 datasets: - BAAI/JudgeLM-100K language: - eu base_model: - orai-nlp/Llama-eus-8B pipeline_tag: text-generation --- HiTZ/judge-eus is a language model designed to evaluate Basque text. It was developed for the [MCG-COLING-2025 Shared Task](), which focused on generating counter-narratives against hate speech using a knowledge-based corpus specifically designed for the task. The model served to evaluate the quality of these counter-narratives, assessing their ability to address and mitigate hate speech effectively. The complete code for task evaluation is available in the [hitz-zentroa/eval-MCG-COLING-2025](https://github.com/hitz-zentroa/eval-MCG-COLING-2025?tab=readme-ov-file) repository.