Time‑series tasks demand both accurate forecasts and rapid decisions. Recent large‑model advances have introduced decision‑oriented models that directly output answers with associated probabilities, referred to here as Structured System 1 models. Their suitability for time‑series problems has not been thoroughly examined. This study uses Jev 1.13.0 to benchmark four tasks: time‑series classification, event‑type prediction, oil‑temperature direction, and stock‑movement prediction. Continuous forecasting targets are discretized into binary or multi‑class decisions to fit the model interface. A unified evaluation framework based on accuracy and AUC is established across tasks. Results reveal strong performance on explicit temporal patterns and event‑type prediction, while many numerical‑sequence classification and stock‑movement tasks hover near chance level. Some datasets also show a gap between ranking quality and final decision accuracy. These findings motivate enhancements in temporal representation, domain adaptation, and decision‑score alignment. The paper additionally supplies standardized sample construction, model invocation, and result‑recording procedures to support future model comparisons.
Review