Abstract page for arXiv paper 2610.28637: To Keep or Not to Keep: Learning KV Cache Retention in Disaggregated LLM Serving Systems