Verified Free-Energy Dynamics: A Hybrid Invariance Principle for Evidence-Gated Persistent Agents

The free-energy principle and active inference describe how an adaptive system can reduce variationalfree energy through perception, learning, and action. They do not, by themselves, require that aproposition admitted to an Agent’s persistent canonical state has passed an external evidence test.An Agent can therefore become internally predictable while retaining unsupported commitments. Weintroduce verified free-energy dynamics (vFEP), an engineering extension that separates homeostaticor inferential coherence from epistemic admissibility. The construction is hybrid: a continuous freeenergy flow regulates the Agent’s internal state, while a discrete verification projection controlswhich candidate records may become canonical. A non-negative verification deficit is added to adimensionless base-energy residual, yielding a verified coherence that cannot be maximal when anyunsupported commitment is canonical. We establish forward invariance of the verified manifold,exponential convergence of the continuous coordinate under an explicit Polyak–Lojasiewicz condition,and preservation of the invariant under atomic, proof-matched self-modification. A controlled gateablation illustrates the central empirical distinction: ordinary coherence can remain near one whileunsupported canonical records accumulate, whereas verified coherence falls immediately; restorationof the gate and certificate re-scan removes the deficit. The claim is deliberately limited: vFEPguarantees compliance with a specified verification policy, not metaphysical truth, verifier infallibility,consciousness, or general intelligence.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-07-16
DOI
https://doi.org/10.5281/zenodo.21400252
Primary Topic
Embodied and Extended Cognition
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
preprint

Verified Free-Energy Dynamics: A Hybrid Invariance Principle for Evidence-Gated Persistent Agents

Steven Domenikos
Zenodo (CERN European Organization for Nuclear Research)
Embodied and Extended Cognition
preprint

Verified Free-Energy Dynamics: A Hybrid Invariance Principle for Evidence-Gated Persistent Agents

Steven Domenikos
preprint en

Abstract

The free-energy principle and active inference describe how an adaptive system can reduce variationalfree energy through perception, learning, and action. They do not, by themselves, require that aproposition admitted to an Agent’s persistent canonical state has passed an external evidence test.An Agent can therefore become internally predictable while retaining unsupported commitments. Weintroduce verified free-energy dynamics (vFEP), an engineering extension that separates homeostaticor inferential coherence from epistemic admissibility. The construction is hybrid: a continuous freeenergy flow regulates the Agent’s internal state, while a discrete verification projection controlswhich candidate records may become canonical. A non-negative verification deficit is added to adimensionless base-energy residual, yielding a verified coherence that cannot be maximal when anyunsupported commitment is canonical. We establish forward invariance of the verified manifold,exponential convergence of the continuous coordinate under an explicit Polyak–Lojasiewicz condition,and preservation of the invariant under atomic, proof-matched self-modification. A controlled gateablation illustrates the central empirical distinction: ordinary coherence can remain near one whileunsupported canonical records accumulate, whereas verified coherence falls immediately; restorationof the gate and certificate re-scan removes the deficit. The claim is deliberately limited: vFEPguarantees compliance with a specified verification policy, not metaphysical truth, verifier infallibility,consciousness, or general intelligence.

Zenodo (CERN European Organization for Nuclear Research)
Criterion Health (United States) (US)
Affordable and clean energy
Embodied and Extended Cognition
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.