Compiler speeds up private AI model queries by 24 to 50 percent

Original: FuseFSS: Efficient Secure LLM Inference with Function Secret Sharing

Source: arxiv.org ↗