We introduce a representation learning framework that provides high-confidence fairness guarantees with controllable error thresholds and confidence levels via adversarial inference.

AdaDetectGPT is all your need for detecting LLM-generated text.