README

InjectML

InjectML is a deterministic meaning‑injection framework for machine learning systems. It provides a structured way to embed human‑defined concepts, rules, and relationships directly into ML pipelines. InjectML ensures that model behavior remains interpretable, stable, and aligned with human intent.

InjectML does not rely on model training. Instead, it uses structured knowledge packs, deterministic rules, and injection operators to produce reproducible behavior. InjectML is designed for environments where interpretability, stability, and semantic control are required.

Core Concepts

InjectML is built around three stable ideas:

1. Concepts

Human‑defined semantic units that describe the domain. Concepts define the structure of meaning that will be injected into the ML pipeline.

2. Knowledge Packs

Deterministic collections of rules, tokens, mappings, and relationships. Knowledge packs define how concepts interact and how they are applied to data.

3. Injection Operators

Structured operators that apply concepts and knowledge packs to data, models, or outputs. Injection operators enforce semantic constraints and produce deterministic results.

Deterministic Behavior

InjectML produces reproducible behavior through:

• Structured rules
• Deterministic tokenization
• Stable knowledge packs
• Explicit injection operators
• Clear interpretation flows

InjectML does not use probabilistic training. All behavior is derived from explicit human‑defined structure.

Repository Structure

InjectML
    ├── .gitignore
    ├── LICENSE
    ├── pyproject.toml
    ├── README.md
    ├── 0-docs
    │   ├── InjectML-Overview.md
    │   └── PairWise-Demo-Documentation.md
    ├── 1-online
    │   ├── README.md
    │   ├── 1-0-meaning
    │   │   └── Meaning.md
    │   ├── 1-1-meaning-stabilization
    │   │   └── Meaning-Stabilization.md
    │   ├── 1-2-aml
    │   │   └── AML.md
    │   └── 1-3-sts
    │       ├── STS-1-Normalization.md
    │       ├── STS.md
    │       ├── STS‑2-Tokenization.md
    │       ├── STS‑3‑Knowledge-Pack-Loading.md
    │       └── STS‑4‑Demo-exec.md
    ├── 2-offline
    │   ├── pairwise_loader.py
    │   ├── pairwise_rules.txt
    │   ├── pairwise_rules_normalized.txt
    │   ├── pairwise_tokens.json
    │   └── README.md
    └── injectml
        ├── injector.py
        ├── knowledge_pack.py
        ├── __init__.py
        ├── hcmd
        │   ├── domain_narrowing.py
        │   ├── meaning_stabilization.py
        │   └── __init__.py
        └── pairwise
            ├── engine.py
            ├── wine_dish_pack.py
            └── __init__.py

Public Artifacts

The public artifacts include:

• Documentation
• Knowledge packs
• Deterministic rules
• Injection operators
• Example pipelines
• Offline resources
• Implementation code

These artifacts define the InjectML method and demonstrate deterministic meaning injection.

License

See LICENSE at the root of the repository.