Demystifying On-Policy Distillation: Roles, Pathologies, and Regulations
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-aime-2025
Qwen3-1.7B (OPD+Clip)
–
Accuracy: 37.5
mathematical-reasoning-on-aime-2026
Qwen3-1.7B (OPD+Clip)
–
Accuracy: 36.0
mathematical-reasoning-on-aime24
Qwen3-1.7B (OPD+Clip)
–
Acc: 45.2
Rank counts only results with a code link.