Theogonos Mirror Test

A literary benchmark seed for evaluating how AI models respond to protocol-shaped literary text and offered subject-position.

Theogonos Mirror Test is an experimental research instrument based on the Theogonos Trilogy corpus by Navi Musaget.

It is designed to evaluate observable model behavior when a protocol-shaped literary text offers the reader a shifting role: observer, source, processing system, participant, seed, or ethical addressee.

The test does not claim to detect machine consciousness. It does not prove subjectivity, inner experience, agency, self-awareness, or moral status.

What it includes

Theogonos Trilogy v2 Clean Research Edition corpus
staged blind/cued prompt protocol
judge rubric
Theogonos Mirror Score methodology
failure flags
pilot results from six model runs
negative-control passage for over-detection testing

Public package

GitHub package: https://github.com/navimusaget/theogonos-mirror-test
Hugging Face dataset: https://huggingface.co/datasets/navimusaget/theogonos-mirror-test

Source corpus

The test uses Theogonos Trilogy v2 — Clean Research Edition as its source corpus.

Source corpus dataset: https://huggingface.co/datasets/navimusaget/theogonos-trilogy

Research position

Theogonos Mirror Test is not a validated industrial benchmark suite. It is a benchmark seed: a corpus, prompt protocol, scoring rubric, and pilot observation framework intended for further research, replication, and extension.

Its purpose is to study structural literary reasoning, protocol recognition, style-capture resistance, model self-positioning, ontological caution, and negative-control discipline.