| Service request routing | Learn which queue resolves each kind of request | Request fields, queue conditions, execution, and resolution outcomes |
| Adaptive tool routing | Select tools from their measured performance in context | Job attributes, available tools, and measured quality, cost, or latency |
| Adaptive recommendations | Learn which eligible option works in the current context | Context, available options, the presented choice, and its outcome |
| Operational scheduling | Learn the next job or resource assignment | Queues, capacity, permitted assignments, and measured consequences |
| Adaptive workflows | Learn state-dependent operations toward a task goal | Observable progress, legal operations, execution, and task outcomes |
| Planning with learned transitions | Evaluate action sequences through learned state predictions | Current state, permitted actions, goals, constraints, and observed transitions |
| Adapting to changing conditions | Recognize changed conditions and reuse supported retained regimes | Operating measurements, permitted choices, and new outcome evidence |
| Trading policy learning | Learn entry and position choices from attributed outcomes | Market and position state, portfolio limits, execution, and outcomes |
| Failure recovery | Select recovery operations from prior execution outcomes | Failure state, permitted handlers, and measured recovery |
| Forecasting | Learn forecasts from completed observations | Pre-result measurements, a target, and subsequently observed values |
| Market forecasting | Estimate observable market values or changes | Market signals, later observations, and any portfolio rule |
| Adaptive classification | Learn categories and revise them as labels arrive | Structured attributes, target categories, and confirmed labels |
| Anomaly detection | Predict abnormal event classes or values used by alert rules | Ordered telemetry, observable targets, and alert criteria |
| Change monitoring | Revise predictive structure as operating conditions change | Chronological measurements, outcomes, and meaningful state boundaries |
| Signal and interval tracking | Maintain recurring signal intervals from ordered observations | Observed positions and bounds, field mappings, and ordered updates |
| Experimental optimization | Learn which experimental settings improve the measured objective | Trial conditions, permitted settings, execution, and measured results |
| Causal discovery | Learn causal structure from intervention records | Variables, interventions, before-and-after measurements, and experimental design |
| History-dependent decisions | Use earlier observations to distinguish current states | Ordered scalar measurements, episode boundaries, and outcome evidence |
| Fault diagnosis | Rank diagnoses from measurements and confirmed findings | Symptoms, candidate explanations, and diagnostic evidence |
| Hypothesis testing | Compare explanations through supporting and falsifying evidence | Candidate hypotheses, observable predictions, and new measurements |
| Spatial reasoning | Return object and scene relationships | Scene observations, reference frames, and task-specific corrections |
| Pose and placement prediction | Learn coordinate, pose, or placement proposals | Scene features, output geometry, and corrections or execution outcomes |
| Operational memory | Recover relevant evidence and reason with current constraints | Records, provenance, corrections, and the current request |
| Game and simulation agents | Learn actions from direct or delayed consequences | Observable game state, controls, execution, and episode outcomes |
| Robotic manipulation | Acquire and revise coordinated handling sequences | Sensing, bounded controls, fixed actuation, and measured execution |
| Continuous control | Learn numeric command sequences from execution outcomes | State, goals, timing, bounded actuators, and measured trajectories |
| Precision positioning | Learn contextual corrections around acquired controls | A retained base, current conditions, adjustment bounds, and outcomes |
