TREK: Distill to Explore, Reinforce to Refine
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-aime-2025
Qwen3-8B + TREK (DeepSeek-V4 proposals)
–
Accuracy: 40.3
mathematical-reasoning-on-aime24
Qwen3-8B + TREK (DeepSeek-V4 proposals)
–
Acc: 51.1
Rank counts only results with a code link.