MTEB eval

[DRRV]

Hi

First thanks for this work. really interesting!

I was wondering how you perform the evaluation for the MTEB(eng, v2) Benchmark for reranking. The code seems to correspond to the retrieval evaluation.
Thanks
Hervé

[liuqi6777]

Hi,

For the MTEB evaluation, we just follow the standard pipeline, including reranking tasks (https://github.com/Alibaba-NLP/E2Rank/tree/main/eval_mteb) That means we use the standard instructions.

If you want to use the listwise prompt in the paper to evaluate the reranking benchmarks, we provide other code and scripts for reference.