Abstract: Existing reinforcement learning-based data indexes optimize structures using fixed objectives, making them unsuitable for dynamic environments with varying resources and latency requirements ...