Towards Privacy-Preserving Large Language Model: Text-free Inference Through Alignment and Adaptation

Jeongho Yoon, Chanhee Park, Yongchan Chun +2

Current LLM-based services typically require users to submit raw text regardless of its sensitivity. While intuitive, such practice introduces substantial privacy risks, as unautho…