Skyfall AI Launches MORPHEUS: Enterprise Simulation Benchmark for Continual Reinforcement Learning

Skyfall AI released MORPHEUS, an enterprise-grade simulation benchmark for continual reinforcement learning. Its core feature is a never-reset world, simulating structured non-stationary environments via parametric regime shifts, with a six-metric evaluation protocol. On this platform, mainstream algorithms like PPO, HER, EWC, and LCM all fall far below theoretical upper bounds, exposing their shortcomings in continual learning. For RL research teams, MORPHEUS provides a testbed closer to real-world continuous operation scenarios, helping evaluate algorithm stability and adaptability under dynamic regime changes. However, current SOTA performance suggests the need for more robust continual learning solutions.