This research addresses "decision-metric alignment" in JEPA-style latent world models, noting that strong task variable decoding does not ensure that Euclidean distance to goal latents accurately ranks action sequences by real task progress. The authors introduce Plan-Real Spearman and CEM-stage Spearman as diagnostics to measure rank agreement between latent and real-world progress. These metrics help evaluate the effectiveness of action-conditioned objectives for Model Predictive Control (MPC) planning.