RoboValue introduces a unified benchmark for evaluating robotic value models across four capability dimensions: task-state understanding, temporal progress monitoring, failure and recovery reasoning, and value consistency. The framework employs diagnostic trajectories and instruction variations through shared interfaces to test whether value judgments accurately reflect task requirements throughout execution. The RoboValue-Dataset comprises 15 simulation and 20 real-world manipulation tasks, providing expert demonstrations for training and annotated trajectories for fine-grained evaluation. Researchers assessed representative value models in zero-shot and one-shot settings under standard conditions and under separate shifts in robot embodiment and environment. Results expose systematic errors in fine-grained instruction grounding, history-dependent progress tracking, and recovery assessment, even among models achieving strong outcome and forward-progress scores. Notably, identical semantic subtasks receive substantially different value gains across valid execution orders, revealing inconsistency in value attribution. These findings pinpoint concrete targets for improved value supervision, while the dataset and public leaderboard enable ongoing assessment of feedback reliability for policy learning and execution monitoring.