voyage-large-2-instruct is an instruction-tuned general-purpose embedding model optimized for clustering, classification, and retrieval. For general retrieval tasks, please use input_type argument, which adds instructions on the backend.
For classification (or clustering) tasks, please set input_type to None and prepend "Classify the text: " (or "Cluster the text: ") to your texts.
For MTEB subtasks, please refer to instruct.json for the instruction. Similar to classification and clustering use cases, please set input_type to None and prepend the instruction to your texts.