Cache-to-Cache: Direct Semantic Communication Between Large Language Models Paper • 2510.03215 • Published Oct 3 • 94