Skip to content

Adding results and description for Tomoro ColQwen3-4B + BM25 + doc MaxP + zerank-2 - #146

Open
shirin2105 wants to merge 1 commit into
illuin-tech:mainfrom
shirin2105:add-tomoro-colqwen3-4b-zerank2
Open

shirin2105 wants to merge 1 commit into
illuin-tech:mainfrom
shirin2105:add-tomoro-colqwen3-4b-zerank2

Conversation

@shirin2105

Copy link
Copy Markdown

This PR adds results of an open-weight pipeline (Tomoro ColQwen3-4B + BM25 + document MaxP + zerank-2) to the ViDoRe V3 pipeline benchmark.

  • TomoroAI/tomoro-colqwen3-embed-4b scores page images with MaxSim.
  • When the corpus is in the query language, BM25 on the page markdown is added.
  • Each page is boosted by the best page score of its document.
  • zeroentropy/zerank-2 then reranks the top 50 pages on their markdown.

No generative model or API is used.

The average nDCG@10 over the 8 English datasets is 0.6874. Search takes 5.0–25.3 s per query on 2 × T4 (16 GB), with one query at a time.

  • The pipeline reads the corpus doc_id column only to group pages by document. The text of doc_id is never given to a model. This is stated in the description file.
  • The CLI writes num_queries only per language, so we copied the English value into overall for the CI check.

The pipeline implementation is available at vidore-v3-tomoro-zerank2.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant