Intrinsic Robot Rewarding: Reusing VLA Representations for Autonomous Evaluation and Policy Improvement | TickerVault