This paper introduces Structural Recursive Viability Dynamics (SRVD), a cross-scale framework describing how complex systems sustain viability through recursive information propagation under energetic constraints. By establishing Effective Information (I), Viability Time (T), and Energy Cost (E) as core variables, SRVD constructs a five-fold recursive architecture to model macroscopic non-equilibrium phase transitions, including boundary dissolution, Decoder (D) drift, and Endogenous Time (Tₚred) collapse. Through linear stability analysis, we derive the critical threshold where a system's Virtual Valuation dynamically decouples from its Objective Viability. This allows us to rigorously map AI "reward hacking" as a self-exciting phase transition driven by Tₚred approaching zero, providing a testable reinforcement learning falsification protocol (Section 9. 2). Axiomatically distinct from classical steady-state inference frameworks, SRVD offers an irreducible dynamical mechanism for the evolutionary instability of complex systems.
liangruifeng (Tue,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: