Gains and Collapse in On-Policy Distillation:A Reinforcement Learning Perspective

Open paper
Benchmark Model Rank Results
mathematical-reasoning-on-aime244B → SFT (warmup)#132Acc: 9.17
mathematical-reasoning-on-aime-20254B → SFT (warmup)#144Accuracy: 10.0

Rank counts only results with a code link.