Paper Detail
Zixian Huang, Kaichen Yang, Xu Huang, Feiyang Hao, Qiming Ge, Bowen Li, He Du, Kai Chen, Qipeng Guo
A widely adopted strategy for model enhancement is to use synthetic data generated by a stronger model for supervised fine-tuning (SFT). However, for emerging reasoning models like Qwen3-8B, this approach often fails to improve reasoning capabilities and can even lead to a substantial drop in performance. In this work, we identify substantial stylistic divergence between teacher generated data and the distribution of student as a major factor impacting SFT. To bridge this gap, we propose a Teacher-Student Cooperation Data Synthesis framework (TESSY), which interleaves teacher and student models to alternately generate style and non-style tokens. Consequently, TESSY produces synthetic sequences that inherit the advanced reasoning capabilities of the teacher while maintaining stylistic consistency with the distribution of the student. In experiments on code generation using GPT-OSS-120B as the teacher, fine-tuning Qwen3-8B on teacher-generated data leads to performance drops of 3.25% on LiveCodeBench-Pro and 10.02% on OJBench, whereas TESSY achieves improvements of 11.25% and 6.68%.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@misc{huang2026how,
title = {How to Fine-Tune a Reasoning Model? A Teacher-Student Cooperation Framework to Synthesize Student-Consistent SFT Data},
author = {Zixian Huang and Kaichen Yang and Xu Huang and Feiyang Hao and Qiming Ge and Bowen Li and He Du and Kai Chen and Qipeng Guo},
year = {2026},
abstract = {A widely adopted strategy for model enhancement is to use synthetic data generated by a stronger model for supervised fine-tuning (SFT). However, for emerging reasoning models like Qwen3-8B, this approach often fails to improve reasoning capabilities and can even lead to a substantial drop in performance. In this work, we identify substantial stylistic divergence between teacher generated data and the distribution of student as a major factor impacting SFT. To bridge this gap, we propose a Teach},
url = {https://huggingface.co/papers/2604.14164},
keywords = {supervised fine-tuning, synthetic data, teacher-student collaboration, style transfer, data synthesis, reasoning models, code generation, LiveCodeBench-Pro, OJBench, code available, huggingface daily},
eprint = {2604.14164},
archiveprefix = {arXiv},
}
{}