Repository navigation
Consider removing Contriever examples from the onboarding docs #2680
Description
Should we remove the optional Contriever examples from the onboarding path, given that Contriever is outdated and BGE-base already serves as the main dense-retrieval example?
The path beginning with MS MARCO passage ranking proceeds through conceptual-framework.md, experiments-nfcorpus.md, conceptual-framework2.md, and conceptual-framework3.md.
Contriever appears in three of these documents:
- experiments-nfcorpus.md: Four optional sections covering indexing, batch retrieval, evaluation, and interactive retrieval, including commands and expected outputs.
- conceptual-framework2.md: Eight optional sections covering index inspection, vector reconstruction, document IDs, document encoding, vector comparison, interactive retrieval, query-document scoring, and manual retrieval. Also mentions Contriever in the recap.
- conceptual-framework3.md: A recap sentence naming Contriever alongside BGE-base.
The proposed cleanup would remove these optional sections and update the recap text, keeping a consistent BGE-base walkthrough for the dense-retrieval lessons. This would simplify onboarding and reduce the number of parallel examples and expected outputs to maintain.
The initial MS MARCO guide and conceptual-framework.md contain no Contriever references.
Activity
lilyjge commentedon Sep 13, 2026
yeah I think that would make sense, if we want we could maybe replace the optional dense exercise with a newer embedding model?