This PR adds evaluation results for different models in the public_rerank_benchmarks directory. Specifically, it includes:
Results for BM25, Rerank v2, and Rerank v3 models on various datasets, including BEIR, Code, Long Context, Multilingual, and Semi-Structured Data evaluations.
Notes specifying the model variants used for the Multilingual evaluations and those used for the other evaluations.
A table with evaluation results for BM25 and embed-multilingual-v3.0 models, including language-specific results and averages for 18 datasets.
Results for Rerank 3 on top of BM25 and embed-multilingual-v3.0, showing improvements in most cases.
This PR adds evaluation results for different models in the
public_rerank_benchmarks
directory. Specifically, it includes: