RafaelUI/

RuLitSearch

A dataset for fine-tuning semantic search and text embedding models on Russian classical literature.

284,275 triplets (anchor, paraphrase, negative) extracted from Russian classical literary texts. Hard negatives were mined using microsoft/harrier-oss-v1-270m.

	Schema




	

Field Description

anchor A passage from a literary text (avg ~900 tokens)

paraphrase A short description of the anchor's content (avg ~44… See the full description on the dataset page: https://huggingface.co/datasets/RafaelUI/RuLitSearch.