ColBERT Datasource

ColBERT datasource relies on the RAGatouille backbone.

ColBERTDS uses ColBERTv2, a model that uses the ColBERT architecture, which represents text as multiple contextualized embeddings rather than reducing all tokens into a single vector. This allows ColBERTv2 to produce higher quality embeddings than all-mpnet-base-v2 for certain tasks, particularly those involving longer text or more complex language structures.