mteb/RiaNewsRetrieval_test_top_250_only_w_correct-v2

RiaNewsRetrievalHardNegatives An MTEB dataset Massive Text Embedding Benchmark News article retrieval by headline. Based on Rossiya Segodnya dataset. The hard negative version has been created by pooling the 250 top documents per query from BM25, e5-multilingual-large and e5-mistral-instruct. Task category t2t Domains News, Written Reference https://arxiv.org/abs/1901.07786 How to evaluate on this task You can evaluate an embedding model on this… See the full description on the dataset page: https://huggingface.co/datasets/mteb/RiaNewsRetrieval_test_top_250_only_w_correct-v2.