Split-and-Denoise: Protect large language model inference with local differential privacy

Open in new window