Eliciting Weak-to-Strong Generalization with On-Policy Reverse Distillation

Benchmark Model Rank Results
mathematical-reasoning-on-hmmt-feb-2025Qwen3-8B + OPRD#23Accuracy: 31.42
mathematical-reasoning-on-olympiadbenchQwen3-8B + OPRD#23Accuracy: 53.26
mathematical-reasoning-on-aime24Qwen3-8B + OPRD#55Acc: 66.92
mathematical-reasoning-on-aime-2025Qwen3-8B + OPRD#77Accuracy: 56.04