Paper

Efficient Omni-modal Large Language Model Inference via Per-Query Modality Selection · ByteTrail