Cognition Lab Embedded Web Simulation

Active Methods:
Step: 0 Optimal Path: -- steps

Metric Definitions

1. Learning Rate: Drops as agents learn. Flat = Zero-Shot (Active Inf) or Planning (Dyna-Q).
2. Path Efficiency: 1.0 is Perfect (Optimal). < 0.1 means wandering.
3. Task Success: Cumulative goals reached. Steep slope = High Reliability.
4. Processing Latency (ms): Cost of "Thinking". Active Inf (0.02ms) is faster than Random Forest (0.34ms).
5. Data Efficiency: % of map seen. Lower % + Success = Smart Exploration.
6. Memory Footprint (KB): RAM usage. Q-Learning is tiny; Dyna-Q/Forest grow over time.
7. Cumulative Free Energy (Nats): Measures Surprise. High (10k+) = RL Agents (Learning via Error). Low (<100) = Active Inference (Predictive Homeostasis).