metadata
datasets:
- spanish-ir/messirve
language:
- es
base_model:
- intfloat/multilingual-e5-large
pipeline_tag: feature-extraction
multilingual-e5-large model fine-tuned on the MessIRve Spanish IR full training set, retrieving hard negatives with BM25 and following the same approach as Wang et al. (2024).
Refer to https://github.com/ftvalentini/MessirveSpanishIR for more details on the training.
Paper: MessIRve: A Large-Scale Spanish Information Retrieval Dataset (EMNLP 2025)