README.md

metadata

license: mit
datasets:
  - unicamp-dl/mmarco
language:
  - af
  - nl
base_model:
  - unicamp-dl/mt5-base-mmarco-v2

mt5-base Reranker Linguistic Transfer (Afrikaans-Dutch)

This is a variation of Unicamp's mt5-base Reranker initially finetuned on mMARCO. Our model is post-trained on Dutch translated to Afrikaans Queries and Dutch Documents from MMARCO/v2.

This model was fine-tuned for the "Improving Low-Resource Retrieval Effectiveness using Zero-Shot Linguistic Similarity Transfer" ECIR2025 paper. The source code for the paper can be found here