arXivAI Research
Pivot-SD: Efficient Self-Distillation for Masked Diffusion Language Models
Pivot-SD：針對遮罩擴散語言模型的高效自我蒸餾框架
Pivot-SD is an efficient offline self-distillation framework for masked diffusion language models that identifies high-impact "pivot" decisions using information-gain, optimizing successful and failed trajectories selectively.
2 min read