GEME — Generative Economy Memory Entity

简单直截，发明本心。
— 陆九渊

A Prism, A Telescope

GEME is a minimal computational system built on three axioms:

1. Competitive merging — new input merges into the closest existing frame, or creates one
2. Adaptive forgetting — frames decay and are pruned; survivors are the most verified
3. Self-referential observation — the system observes its own frame economy and feeds back the result

Operating with only three structural constants (δ, γ, τ) from which all behavioral thresholds derive. Zero free parameters. Six emergent cognitive layers (L1–L6). Reproducible in ~10 seconds.

Quick Start

git clone https://github.com/JackeyLGene/GEME.git
cd GEME/submit/code
python s1_demo.py

No external dependencies. Output: the six-layer discovery journey.

Key Results

Claim	Result
Self-reference is informationally cheap	I(φ;X) = 0.026 bits, t(19)=65.2, p<.001
Self-reference ≈ Induction	Q+G ≡ PA (10 seeds, std = 0)
Frame attractor	L4 = 1 ± 0.2, System = 6 ± 2
Six cognitive layers	L1 Entity → L6 Integration

Documentation

• Paper (docx) — full manuscript
• Landing Checklist — pre-submission review
• Code — 980 lines, zero dependencies
• Replication — run Q+G≈PA in 30s
• Essays — short reflections

Citation

@software{liu2026geme,
  author = {Liu, Jieqi},
  title = {GEME: A Self-Referential Prism for Cognitive Modeling},
  month = may,
  year = 2026,
  publisher = {Zenodo},
  doi = {10.5281/zenodo.20110147},
  url = {https://doi.org/10.5281/zenodo.20110147}
}

License

Apache 2.0. © 2026 Jieqi Liu.