CITATION.cff

cff-version: 1.2.0
title: "ARC-Neuron LLMBuilder: A Governed Local AI Build-and-Memory System"
authors:
  - family-names: Doman
    given-names: Gary
    affiliation: Independent
message: "If you use this repository, please cite the project metadata in this file."
abstract: >-
  ARC-Neuron LLMBuilder is a local-first, governed cognition lab that treats a
  language model as one artifact inside a lawful promotion lifecycle. It ships
  a canonical conversation pipeline, a regression-aware promotion gate (Gate v2),
  a floor model for never-below baselines, draft-critique-revise reflection,
  automatic terminology absorption from conversation, an indexed Omnibinary
  ledger with O(1) event lookup, and Arc-RAR restorable archives — all wrapping
  a replaceable model adapter. The v2.0.0-audited release documents post-audit Gate v2 promotion evidence,
  the current `arc_governed_v10_wave4` incumbent, and a regression-aware
  local proof loop for governed model growth.
license: MIT
version: v2.0.0-audited
date-released: 2026-05-04
repository-code: "https://github.com/GareBear99/ARC-Neuron-LLMBuilder"
url: "https://github.com/GareBear99/ARC-Neuron-LLMBuilder"
keywords:
  - local-ai
  - governed-ai
  - cognition-engine
  - promotion-gate
  - regression-floor
  - omnibinary
  - arc-rar
  - gguf
  - benchmarking
  - reflection-loop
  - terminology-absorption
  - conversation-pipeline
  - agentic-ai
  - small-language-models
  - transformer
  - pytorch