Recursive, Containment-Aware Transformer Core
This is the reference implementation of the CVMP Transformer
, an original transformer architecture designed with symbolic routing, recursive containment logic, and dynamic self-healing feedback systems.
Author: Garret Sutherland
Signature: MirrorEthic::Containment_First
License: MIT / Open Containment License
- ORC Router: Symbolic tier-based module routing
- StretchField FFN: DPS-aware transformation
- RISL: Reflective Inertia Smoothing Layer
- BloomCatch: Recursive pattern detection + soft reset
- TierDriftMonitor: Entropy-aware drift tracking
- LogBleed Regularizer: Repetition compression penalty
python3 cvmp_transformer_v1_4r.py
## Benchmark
``bash
python benchmark_runner.py
---
/cvmp-transformer-core
├── cvmp_transformer_v1_4r.py # Full source
├── benchmark_runner.py # Benchmark against base/works with v1.6x
├── cvmp_transformer_v1.6x.py # Newest [status]: Running
├── LICENSE # CVMP-specific open license
├── README.md # YOU ARE HERE
├── test/
│ └── test_demo_output.txt # Optional sample debug output
└── docs/
├── architecture.md # Component descriptions
└── containment_theory.md # CVMP theory primer