AI RESEARCH
PACED: Distillation at the Frontier of Student Competence
arXiv CS.AI
•
ArXi:2603.11178v1 Announce Type: new Standard LLM distillation wastes compute on two fronts: problems the student has already mastered (near-zero gradients) and problems far beyond its reach (incoherent gradients that erode existing capabilities). We show that this waste is not merely intuitive but structurally inevitable: the gradient signal-to-noise ratio in distillation provably vanishes at both pass-rate extremes.