Retro*: Optimizing LLMs for Reasoning-Intensive Document Retrieval Paper • 2509.24869 • Published Sep 29 • 1