Tag
PragAlign introduces a feedback-guided framework for controlled synthetic dialogue generation using an LLM-based evaluator to improve alignment with intent, emotion, coherence, and fluency, achieving 99.50% acceptance compared to 72.25% for one-shot generation.