FSPO: Few-Shot Preference Optimization of Synthetic Preference Data in LLMs Elicits Effective Personalization to Real Users Paper • 2502.19312 • Published 2 days ago • 4 • 2