# 师生协作合成学生一致性SFT数据的框架

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-03-23 08:00
- AIHOT 链接：https://aihot.virxact.com/items/cmo2difm902cdslbadb6ij09p
- 原文链接：https://arxiv.org/abs/2604.14164

## AI 摘要

针对使用强模型合成数据微调推理模型时出现的性能下降问题，本文提出师生协作数据合成框架TESSY。该框架通过让师生模型交替生成风格与非风格标记，使合成数据兼具教师的高级推理能力与学生风格一致性。实验以GPT-OSS-120B为教师、Qwen3-8B为学生，在代码生成任务中，传统方法使LiveCodeBench-Pro和OJBench性能分别下降3.25%和10.02%，而TESSY实现11.25%和6.68%的显著提升。

## 正文

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%.