Synthetic (query → passage) training pairs spanning the financial/legal document world —
NDAs, investment management agreements, credit facilities, M&A agreements, fintech/SaaS contracts,
derivatives, securities, real estate finance, and regulatory filings. Built to train asymmetric
retrieval embedding models (a short natural-language query must retrieve the long passage that
answers it), which is where domain models trained on… See the full description on the dataset page:
https://huggingface.co/datasets/sakshamio/financial-legal-synthetic-retrieval.
