{
  "field_id": "F046",
  "field": "conformal prediction",
  "directions": [
    {
      "direction_id": "G0",
      "name": "Coverage-calibrated predictive sets and intervals",
      "scope": "Construct and evaluate prediction sets or intervals with coverage guarantees for model predictions, including general-purpose conformal calibration, local calibration for covariate or response heterogeneity, and the use of predictive distributions, representations, classifier information, or region geometry to improve the usefulness of the sets."
    },
    {
      "direction_id": "G1",
      "name": "Predictive guarantees for dependent and evolving data",
      "scope": "Establish or maintain conformal prediction guarantees under temporal dependence, sequential arrival, or changing observation conditions, including dependence-aware validity, sequential recalibration, and joint prediction regions across a forecasting horizon."
    },
    {
      "direction_id": "G2",
      "name": "Conformal loss and decision-risk control",
      "scope": "Calibrate predictive outputs, sets, or decisions to guarantee bounds on expected loss, error risk, or decision cost, rather than only guaranteeing that the true outcome falls within a prediction set."
    }
  ]
}