Paper IC-Cache: Efficient Large Language Model Serving via In-context Caching got accepted to SOSP’25