An instruction-tuned lightweight embedding model optimized for following retrieval instructions and task-specific queries.