Resources

Code, datasets, and other resources

Code Repositories

  • CATS metric: A metric for measuring the suitability of a linguistic context.
  • Team GPLSI Task 10: The code for training our models for the SemEval-2025 Task 10 challenge.

Datasets

  • COCOTEROS: A Spanish corpus with contextual knowledge for Natural Language Generation. It is designed for constrained text generation tasks where models must generate coherent phrases based on keywords and linguistic context.
  • MULTICOM: A multilingual commonsense generation benchmark extending COCOTEROS to English, Spanish, Dutch, and Valencian. It tasks models with generating commonsensical sentences including a given triplet of words.