Multi-Agent Behavioral Authentication Benchmark and Reproducibility Artifact

This artifact contains the finalized synthetic benchmark corpus, derived decision windows, validation manifests, result artifacts, curated replay code, Colab/GPU result archives, Raspberry Pi 5 profiling outputs, human quality-audit materials, deployment-boundary diagnostics, and manuscript source/PDF for the accompanying paper on edge-oriented behavioral authentication in multi-agent LLM systems.

Version 1.0.6 refreshes the reproducibility package with venue-neutral paper, code, experiment, and artifact paths. The reviewer-facing archive now stages edge-efficiency materials under artifacts/edge_efficiency/, scripts under code/edge_efficiency/, and RP5 power protocols under experiments/rp5_power/. It also includes paper-local artifact mirrors so the manuscript can be rebuilt from the extracted package.

The package is designed for reproducible evaluation replay rather than bit-identical regeneration of the synthetic conversations. The archived corpus is the citable benchmark object. Scenario planning and turn generation used GPT-4.1, whose outputs may change over time; generation scripts are included as provenance and extension tools.