Pareto-Guided Teacher Alignment for Fair Personalized Text Generation

The paper presents a Pareto-guided teacher alignment framework for personalized text generation that addresses demographic disparities while maintaining personalization fidelity. Key components include revision-based candidate generation, pair-aware feasibility gating, and Pareto-style candidate selection, evaluated on climate change and vaccination persuasion tasks across a controlled demographic grid. The findings indicate that no single alignment strategy excels across all fairness and personalization objectives, underscoring the need for multi-audit model selection in fairness-sensitive applications.