metadata
language: en
tags:
- dense-retrieval
- knowledge-distillation
datasets:
- ms_marco
Margin-MSE Trained ColBERT encoder model only
We provide the encoder model of the complete retrieval trained DistilBert-based ColBERT model. This model is trained with Margin-MSE using a 3 teacher BERT_Cat (concatenated BERT scoring) ensemble on MSMARCO-Passage, for more details check the full model card.
This encoder-only model is used as the oracle for distilling term topic embeddings in our ECIR'23 paper.