MockingBERT: Making Transformer Models Resilient to Adversarial Misspellings

Code to generate and extend the TCAB dataset.

🛡️ Detect and combat fake reviews in multiple languages with RevGuard NLP, leveraging adversarial robustness and explainable AI for real-time analysis.

End-to-end NLI classifier on the Adversarial NLI (ANLI) Round 2 benchmark — EDA, baselines, DeBERTa-v3 evaluation (base & large), error analysis, FastAPI inference API with live model switching, Docker deployment, and interactive web UI.