Our paper on learning when to continue search in multi-round RAG is presented at SIGIR 2025.