Flip-Flop Consistency: Unsupervised Training for Robustness to Prompt Perturbations in LLMs
Flip-Flop Consistency (F2C) is an unsupervised training method that makes Large Language Models more robust to different ways of asking questions or prompts. It works by finding a consensus among various prompt variations and aligning the model's internal representations based on this majority vote, significantly boosting agreement and performance without needing human labels.