Preprint / scholarly article

Semantic Phase Dynamics and Active Inference as a Non-Markovian Open-System Process under Energy and Memory Budgets

K. Takahashi

Published
DOI
10.5281/zenodo.18081679

Full text PDF (Zenodo)

Abstract

The preprint defines semantic phases as history-indexed predictive equivalence classes under energy and finite-memory budgets, and builds an auditable non-Markovian framework for intervention selection, irreversibility via path-space KL, and active inference under telemetry contracts.

Keywords

  • semantic phase dynamics
  • active inference
  • non-markovian
  • open systems
  • predictive equivalence
  • intervention library
  • path-space KL
  • energy budget
  • finite memory
  • telemetry contracts
  • auditability

Identifiers and source records