We introduce InForage, an RL framework that enables LLMs to perform adaptive, multi-step retrieval by rewarding informative intermediate steps.