This is a fine-tuned Sentence-BERT model built on top of LegalBERT to semantically identify and score risky clauses in legal documents. It was trained using legal clause pairs and risk annotations, and is optimized for clause-level risk analysis in contracts and compliance workflows.