The Hallucination Tax of Reinforcement Finetuning | Read Paper on Bytez