RL Convergence vs Judgment Discovery Simulator

Vulnerability Manifold Network Step: 0

Standard RL Path

Judgment Path

Zero-Day RCE

Simulation Controls

Manifold & Strategy Preset

RL Exploitation Bias (γ) 0.85

Abandonment Threshold 0.65

Revisit Memory Depth 5 steps

Judgment Decision Ledger 0 decisions

Click "Run Simulation" or interact with nodes to populate judgment trajectory.

Autopoiesis Thesis: Standard RL gradient steps optimize local expected reward, forcing early convergence into local attractors (Step ~14). The Judgment Agent executes Abandon gestures upon diminishing entropy return, leaping across reward gaps to discover hidden Zero-Day RCE nodes.