TRIGNUM-300M

<div align="center">

🧲 TRIGNUM-300M

The Pre-Flight Check for Autonomous AI

"You wouldn't let a plane take off without a pre-flight check.
Why are we letting AI agents act without one?"

<img src="assets/roadmap_architecture.jpg" width="800" alt="TRIGNUM-300M Architecture Flowchart" /> </div>

---

<div align="center"> <!-- TODO: Add your demo GIF here! 1. Record demo/index.html with ScreenToGif 2. Save as assets/trignum_demo.gif 3. Uncomment line below: --> <!-- <img src="assets/trignum_demo.gif" width="800" alt="TRIGNUM-300M Demo" /> --> </div>

What Is This?

TRIGNUM-300M is a zero-model reasoning integrity validator for LLM outputs. It catches structural logic failures — contradictions, circular reasoning, non-sequiturs — before an AI agent acts on them.

from trignum_core.subtractive_filter import SubtractiveFilter

sf = SubtractiveFilter()
result = sf.apply(agent_output)

if result.illogics_found:
    agent.halt(reason=result.illogics_found)
    # T-CHIP glows RED 🔴 → Human review required
else:
    agent.execute()
    # T-CHIP glows BLUE 🔵 → Cleared for takeoff

No LLM. No API. No training data. ~300 lines of Python. <1ms.

---

🔬 Benchmark Results

We expanded our evaluation to 58,000+ real LLM outputs including a new 517-sample curated dataset for structural reasoning. Honest results:

What this means:

• 99.5% F1 on structural reasoning failures — contradictions, circular logic, unsupported conclusions
• 4.0% F1 on factual hallucinations — we don't catch wrong facts

That's the point. There are 100 tools for fact-checking. There are zero tools for reasoning-checking. Until now.

Per-Task Breakdown (HaluEval)

Throughput: 146,866 samples/second — orders of magnitude faster than LLM-based validation.

---

✈️ The Pre-Flight Check Analogy

A pre-flight checklist doesn't verify that London exists. It verifies that:

• ✅ Instruments don't contradict each other
• ✅ There are no circular faults (sensor A confirms B confirms A)
• ✅ The flight computer draws conclusions from actual data
• ✅ Systems are logically consistent

The Subtractive Filter does the same for AI reasoning:

LLM Output → Subtractive Filter → [PASS] 🔵 → Agent Executes
                                 → [FAIL] 🔴 → Agent Halts → Human Review

---

🤖 The Missing "Agentic Validator"

In the context of the recent shift towards Agentic Reasoning, autonomous LLMs are moving from static prompts to dynamic thought-action loops involving planning, tool-use, and multi-agent collaboration.

Current systems rely heavily on probabilistic models to act as the "Critic/Evaluator" or use "Validator-Driven Feedback" via unit tests for code or simulators for robotics. But there has been no validator for pure logic. If an agent hallucinates a non-sequitur or circular justification during its internal planning phase, the error cascades.

TRIGNUM-300M fills this exact gap. It acts as a deterministic, <1ms Validator-Driven Feedback gate. It halts execution if the agent's internal thought (zt) contains a structural illogic, providing an immediate failure signal (rt = 0) before the agent commits to an irreversible external action (at).

---

🔺 Core Architecture

The Trignum Pyramid

Three faces acting as magnetic poles for data separation:

T-CHIP: The Tensor Character

╔═══════════════════════════════════════════════════════╗
║  T-CHIP [v.300M]                                      ║
║                                                       ║
║  🔵 Blue  = Logic Stable (Cleared for Takeoff)        ║
║  🔴 Red   = Illogic Detected (THE FREEZE)             ║
║  🟡 Gold  = Human Pulse Locked (Sovereign Override)   ║
║                                                       ║
║  Response time: <1ms | False alarms: 0% (structural)  ║
╚═══════════════════════════════════════════════════════╝

The Subtractive Filter

Four detection layers, all pattern-based:

---

📦 Repository Structure

TRIGNUM-300M-TCHIP/
├── src/
│   └── trignum_core/              # Core Python library
│       ├── pyramid.py             # Trignum Pyramid (3 magnetic faces)
│       ├── tchip.py               # T-CHIP (glow states)
│       ├── subtractive_filter.py  # ★ The Subtractive Filter
│       ├── human_pulse.py         # Human sovereignty layer
│       └── magnetic_trillage.py   # Data separation
├── tests/                         # 34 unit tests (all passing)
├── benchmarks/
│   ├── hallucination_benchmark.py     # Curated structural test
│   ├── full_halueval_benchmark.py     # Full 58K HaluEval test
│   ├── results.json                   # Structural benchmark results
│   └── full_halueval_results.json     # Full HaluEval results
├── demo/
│   └── index.html                 # Three.js 3D interactive demo
├── paper/
│   └── TRIGNUM_300M_Position_Paper.md  # Position paper
├── docs/
│   └── theory/                    # 6 foundational theory documents
├── T-CHIP CLEARED FOR TAKEOFF.md  # The pitch
└── ROADMAP.md                     # 2-quarter development plan

---

🚀 Quick Start

# Clone
git clone https://github.com/trace-on-lab/trignum-300m.git
cd trignum-300m

# Install
pip install -r requirements.txt
pip install -e .

# Run the structural benchmark
python benchmarks/hallucination_benchmark.py

# Run the full HaluEval benchmark (downloads ~13MB of data)
python benchmarks/full_halueval_benchmark.py

# Run tests
pytest tests/ -v

---

🌐 Prior Art: Nobody Is Doing This

We searched arXiv, ResearchGate, ACL Anthology, and Semantic Scholar. Every existing reasoning validation system requires model inference:

Existing work uses LLMs to check LLMs. TRIGNUM uses logic to check LLMs.

Read the full analysis in our position paper.

---

⚛️ Quantum Integration: TQPE

TRIGNUM-300M serves as Phase 1 ("Technical A Priori Validation") for Trignumental Quantum Phase Estimation (TQPE).

In our groundbreaking case study estimating the ground state energy of the H₂ molecule, TRIGNUM successfully validated the physical consistency and structural logic of the quantum circuit before execution. By acting as the preliminary gatekeeper, TRIGNUM ensured that no quantum resources were wasted on structurally ill-formed configurations, enabling an epistemic confidence score of 82.8% on the final estimate (-1.1384 Ha).

Read the full BUILDING THE BRIDGE paper on Trignumentality and TQPE in the foundational Trignumentality repository.

---

📚 Documentation

---

💎 The Golden Gems

---

🤝 Contributing

See CONTRIBUTING.md for guidelines.

---

📄 License

MIT License — see LICENSE.

---

📞 Contact

TRACE ON LAB
📧 [email protected]

---

🛡️ The Call

"The most dangerous AI failure is not a wrong fact. It is reasoning that sounds right but isn't."

╔═══════════════════════════════════════════════════════╗
║  🧲 TRACE ON LAB — TRIGNUM-300M — v.300M              ║
║                                                       ║
║  The Pre-Flight Check for Autonomous AI.              ║
║  Zero models. Zero API calls. 146,866 samples/second. ║
║                                                       ║
║  🔵 T-CHIP: CLEARED FOR TAKEOFF.                      ║
╚═══════════════════════════════════════════════════════╝

⭐ Star this repo if you believe AI should check its logic before it acts.