**Problem**: Embedding similarity alone is weak for the final decision. **Task**: Add Reranker module after candidate generation. **What to implement**: - Cohere Rerank integration - Cross-encoder SBERT option - Score normalization Reference: https://github.com/sciknoworg/OntoAligner/issues/77#issuecomment-3855932106
Problem: Embedding similarity alone is weak for the final decision.
Task: Add Reranker module after candidate generation.
What to implement:
Reference: #77 (comment)