Waypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy Distillation

Benchmark Model Rank Results
mathematical-reasoning-on-aime-2026Waypoint-OPD (Qwen3-14B)#43Accuracy: 63.75
mathematical-reasoning-on-aime24Waypoint-OPD (Qwen3-8B)#52Acc: 68.85
mathematical-reasoning-on-aime-2025Waypoint-OPD (Qwen3-14B)#69Accuracy: 60.52