Back to papers
May 6, 2026cs.LGcs.AI

Preference-Based Self-Distillation: Beyond KL Matching via Reward Regularization

Categories

cs.LG, cs.AI