[2410.23214] Grounding by Trying: LLMs with Reinforcement Learning-Enhanced Retrieval