Model Card for Qwen3-1.7B-Open-R1-ADPO

This model is a fine-tuned version of Qwen/Qwen3-1.7B on the watermelonhjg/MATH-lighteval-level_3 dataset. It has been trained using TRL.

Quick start

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="wzx111/Qwen3-1.7B-Open-R1-ADPO", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with ADPO, a method introduced in Anchored Direct Preference Optimization.

Framework versions

• TRL: 0.26.0.dev0
• Transformers: 4.57.1
• Pytorch: 2.9.0+cu126
• Datasets: 4.4.1
• Tokenizers: 0.22.1

Citations

Cite ADPO as:

@misc{zixian2025adpoanchoreddirectpreference,
    title={ADPO: Anchored Direct Preference Optimization}, 
    author={Wang Zixian},
    year={2025},
    eprint={2510.18913},
    archivePrefix={arXiv},
    primaryClass={cs.LG},
    url={https://arxiv.org/abs/2510.18913}, 
}

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}