library_name: transformers
pipeline_tag: text-classification
TESSY: Boundary Predictor
This model is a boundary predictor developed as part of the TESSY (Teacher–Student Cooperation Data Synthesis) framework, as presented in the paper: How to Fine-Tune a Reasoning Model? A Teacher–Student Cooperation Framework to Synthesize Student-Consistent SFT Data.
Model Description
TESSY is a framework designed to bridge the stylistic gap between teacher-generated synthetic data and the student model's distribution. It interleaves teacher and student models to alternately generate style and non-style tokens.
The Boundary Predictor is a critical component of this system; it identifies the transitions between reasoning steps (capability tokens) and non-reasoning stylistic content (style tokens) within a given problem. This allows the framework to determine when to switch between the teacher model (for reasoning) and the student model (for style).
Links
- Paper: arXiv:2604.14164
- Code: GitHub - CoopReason/TESSY
- Dataset: TESSY-Code-80K
Citation
If you find this work useful, please cite:
@article{TESSY,
title={How to Fine-Tune a Reasoning Model? A Teacher--Student Cooperation Framework to Synthesize Student-Consistent SFT Data},
author={Huang, Zixian and Yang, Kaichen and Huang, Xu and Hao, Feiyang and Ge, Qiming and Li, Bowen and Du, He and Chen, Kai and Guo, Qipeng},
journal={arXiv preprint arXiv:2604.14164},
year={2026}
}