FactCheckAI ML Ensemble Server

Fine-tuned RoBERTa models for fake news detection.

  • Model A: 96.3% accuracy (daniB2112 dataset)
  • Model B: 79.8% accuracy (mixed datasets)
  • Model C: DeBERTa fallback
  • Ensemble: Weighted average (0.6/0.4)

API Endpoints:

  • GET /health - Server status
  • POST /predict - Classify claim
Examples