Prediction accuracy
A scalar measuring how closely a world-model forecast matched an observed outcome: 1 - error / scale where scale = max(1, |expected|). A latency prediction of 50ms against an observed 1200ms produces an accuracy score of 0.000, which feeds directly into world-model correction.
Prediction accuracy is computed by the grounding engine after every world-model prediction that has an observable outcome. The formula 1 - error / scale normalizes the absolute prediction error by the magnitude of the expected value, producing a score between 0 and 1 regardless of the unit or scale of the prediction. The scale floor of max(1, |expected|) prevents division by near-zero expected values from producing nonsensical accuracy scores when the world model predicts a near-zero outcome. An accuracy score of 1.0 means the prediction was perfect. A score of 0.5 means the error was half the magnitude of the expected value. A score of 0.0 means the error equaled or exceeded the expected magnitude - a complete miss. These scores feed back into the world model to adjust its confidence estimates: a domain where predictions consistently score below 0.5 causes the model to lower its confidence for future predictions in that domain, which affects arbitration (which uses world-model confidence as a 30% weight) and triggers the curiosity engine to investigate the high-uncertainty domain. Prediction accuracy is also a component of the reinforcement scoring function (14% weight), connecting world-model accuracy improvement directly to memory reward.