How to Fine-Tune a Reasoning Model? A Teacher-Student Cooperation Framework to Synthesize Student-Consistent SFT Data

Benchmark Model Rank Results
code-generation-on-livecodebench-v5TESSY#10Pass@1: 62.87
mathematical-reasoning-on-aime24TESSY#29Acc: 80.42
code-generation-on-livecodebench-v6TESSY#46Pass@1: 55.43
mathematical-reasoning-on-aime-2025TESSY#57Accuracy: 70.1