Self-Distillation

• SDFT: Self-Distillation Enables Continual Learning
• SDPO: Reinforcement Learning via Self-Distillation
• Aligning Language Models from User Interactions