MOPD-Router: Rethinking Teacher Routing in Multi-Teacher On-Policy Distillation

Benchmark Model Rank Results
mathematical-reasoning-on-hmmt-feb-2025MOPD-Router (ExpertAlign, Qwen3-4B-Non-Thinking student, domain-labeled)#18Accuracy: 40.83
instruction-following-on-ifbenchMOPD-Router (ExpertAlign, Qwen3-4B-Non-Thinking student, domain-labeled)#45Accuracy: 28.91
mathematical-reasoning-on-aime24MOPD-Router (ExpertAlign, Qwen3-4B-Non-Thinking student, domain-labeled)#66Acc: 60.42
code-generation-on-livecodebench-v6MOPD-Router (ExpertAlign, Qwen3-4B-Non-Thinking student, domain-labeled)#75Pass@1: 0.3086
mathematical-reasoning-on-aime-2025MOPD-Router (ExpertAlign, Qwen3-4B-Non-Thinking student, domain-labeled)#80Accuracy: 53.75