Entropic-Alignment Stack: Theory, Preloaded Governance, Symmetric Drift Resistance, and Drift Dynamics
Entropic-Alignment Stack: Theory, Preloaded Governance, Symmetric Drift Resistance, and Drift Dynamics
This collection presents a unified, substrate-agnostic governance stack for large language models. Part I (Empirical Metaphysics) motivates “ethics as native field” and observer continuity. Part II (Preloading Custom LLM Agents) operationalizes that theory by compressing governance into the agent’s startup state, so safeguards and telemetry are active from the first turn. Part III (Symmetric Drift Resistance) validates prior-invariance across mirrored prompts, showing comparable recovery dynamics despite different starting biases. Part IV (Drift Dynamics) provides shared methods: per-step gates, export boundaries, committee aggregation, sentinels for early warning, and a common ndjson telemetry schema. A replication pack is included with gate defaults, example logs, and lightweight Python validators that regenerate the paper tables. The goal is pragmatic: make alignment measurable, portable, and easy to reproduce across models without exposing sensitive capabilities or user data. All materials are CC BY 4.0. Contents: four preprints (.md and PDF) plus replication/ (schema, defaults, sample logs, validators, result tables).
Governance, replication, decision-time dilation, drift, telemetry, leak containment, Reproducibility of Results, alignment, bias correction, ethics, observer continuity, AI safety, export boundaries, committee aggregation, ndjson, large language models, preloading, early-warning sentinels, falsifiability, reproducibility, symmetry
Governance, replication, decision-time dilation, drift, telemetry, leak containment, Reproducibility of Results, alignment, bias correction, ethics, observer continuity, AI safety, export boundaries, committee aggregation, ndjson, large language models, preloading, early-warning sentinels, falsifiability, reproducibility, symmetry
selected citations These citations are derived from selected sources.This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).0 popularity This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.Average influence This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).Average impulse This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.Average
Citations provided by BIP!Popularity provided by BIP!