Test any text through the 14-layer governance pipeline.
Explore how text maps to the 6 Sacred Tongues. Each tongue is a domain dimension weighted by phi (golden ratio).
H(d, R) = R^(d^2) — the exponential cost function. As distance from the safe center increases, adversarial cost grows exponentially.
SCBE-AETHERMOORE Governed Training Data
Repository: issdandavis/scbe-aethermoore-training-data
8 Data Views — 394,153 Total Records
| View | Records | Format | Use Case |
|---|---|---|---|
| canonical | 30,949 | FU-structured master | Full pipeline training |
| trl_conversation | 20,949 | Multi-turn messages | TRL SFTTrainer |
| trl_prompt_completion | 20,949 | Prompt + completion | TRL basic |
| openai_chat | 20,949 | OpenAI messages format | GPT-compatible fine-tuning |
| activation_cls | 30,949 | Classification messages | Tongue activation classifier |
| governance_cls | 30,949 | Classification messages | Governance decision classifier |
| contrast_pairs | 2,237 | Positive/negative pairs | DPO / preference training |
| station | 236,222 | Full operational data | Production training runs |
Governance Coverage
Every record is governed — audited through the 14-layer pipeline before publication.
- ALLOW: Normal operation records
- QUARANTINE: 25 records flagged for review (station view only)
- 0 DENY records in any published view
Schema: Canonical Master
{
"substrate": {"type": "...", "source": "..."},
"activation": {"tongue": "KO", "strength": 0.85},
"relation": {"target": "...", "type": "..."},
"permission": {"class": "ALLOW", "governance_hash": "..."},
"flow": {"direction": "inbound", "priority": 3},
"target": "...",
"fu_status": {"active": true, "confidence": 0.92}
}
Quick Start
from datasets import load_dataset
# Load any view
ds = load_dataset("issdandavis/scbe-aethermoore-training-data", "openai_chat")
# Use with TRL
ds = load_dataset("issdandavis/scbe-aethermoore-training-data", "trl_conversation")
# Full canonical master
ds = load_dataset("issdandavis/scbe-aethermoore-training-data", "canonical")
Training Compatibility
| Framework | View | Notes |
|---|---|---|
| TRL SFTTrainer | trl_conversation | Direct — set dataset_text_field="text" |
| OpenAI fine-tune | openai_chat | Export messages column |
| Axolotl | trl_conversation | Use conversation format |
| DPO/RLHF | contrast_pairs | Positive/negative pairs |
| Custom classifier | activation_cls, governance_cls | Classification labels in messages |
All data governed under SCBE-AETHERMOORE 14-layer pipeline. Patent pending USPTO #63/961,403.
Help Train Our Models
Disagree with a governance decision? Tell us. Your feedback directly improves the next training run.
Multi-Model Round Table
Ask a question and get simultaneous responses from multiple AI models, each assigned a Sacred Tongue role. Toggle web search for grounded answers.
SCBE-AETHERMOORE | Built by Issac Davis | Patent Pending USPTO #63/961,403
Every interaction on this page is logged as training data under SCBE governance. By using this tool, you contribute to building safer AI systems.