Fetching the paper…

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