mteb/NQ_test_top_250_only_w_correct-v2

NQHardNegatives An MTEB dataset Massive Text Embedding Benchmark NFCorpus: A Full-Text Learning to Rank Dataset for Medical Information Retrieval. 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 None Reference https://ai.google.com/research/NaturalQuestions/ How to evaluate on this task You can evaluate an embedding model on… See the full description on the dataset page: https://huggingface.co/datasets/mteb/NQ_test_top_250_only_w_correct-v2.