Clearing Up Misconceptions About Chinese Models Relying on Distillation

i_dg23 · x · 2026-07-19

Addressing recent claims that "Chinese models score high only because of distillation," the original poster clarified that those who take distillation technology seriously do not believe Chinese models' progress relies solely on it. Chinese model teams are also doing their own reinforcement learning (RL); distillation just helps them kickstart and accelerate the process. Skeptics might be ignoring the technical insights brought by R1 and R1-Zero.

Original post →

More from Models

Models channel →