NeurAlchemy Threat Matrix Analyzer

Analyze any text input against all NeurAlchemy security models simultaneously:

• Binary DistilBERT — Is it benign or malicious? (99.1% accuracy)
• Multiclass DistilBERT — Which of 7 threat classes? (80.9% accuracy)
• 6 Expert Models — One-vs-rest specialists vote on specific threats
• 4 Classical ML Models — TF-IDF baselines for comparison

Enter any prompt, instruction, or text below to see how each model classifies it.

Models: All models on HuggingFace | Paper: AITL on Zenodo | Code: GitHub