This model is a fine-tuned version of Qwen/Qwen2.5-7B-Instruct. It has been trained using TRL. The reward signal during GRPO training was provided by joke-rater-roberta-zh.

Quick start

python

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="KonradBRG/Qwen2.5-7B-Instruct-Jokester-Chinese", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

Visualize in Weights & Biases

This model was trained with GRPO, a method introduced in DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models.

Framework versions

TRL: 0.23.1
Transformers: 4.57.3
Pytorch: 2.9.1
Datasets: 4.4.1
Tokenizers: 0.22.1

Citations

If you use this model, please cite the paper it was built for:

Team TüLK at SemEval-2026 Task 1: Humor Generation with Qwen and Group Relative Policy Optimization

bibtex

@inproceedings{bruggemann-hou-2026-team,
    title = {Team {T}{\"u}{LK} at {S}em{E}val-2026 Task 1: Humor Generation with Qwen and Group Relative Policy Optimization},
    author = {Br{\"u}ggemann, Konrad  and
      Hou, Luting},
    editor = "Kochmar, Ekaterina  and
      Ghosh, Debanjan  and
      North, Kai  and
      Komachi, Mamoru",
    booktitle = "Proceedings of the 20th {I}nternational {W}orkshop on {S}emantic {E}valuation (2026)",
    month = jul,
    year = "2026",
    address = "San Diego, California, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.semeval-1.67/",
    doi = "10.18653/v1/2026.semeval-1.67",
    pages = "463--474",
    ISBN = "979-8-89176-414-9",
    abstract = "This paper addresses the challenge of computational humor generation proposed in SemEval-2026 Task 1: Humor Generation. Our approach leverages Group Relative Policy Optimization, with an LLM serving as the policy and a custom joke rating model providing a reward signal. We demonstrate that this framework is an effective and computationally efficient approach, reliably producing genuinely funny content that adheres to task constraints."
}

Cite GRPO as:

bibtex

@article{shao2024deepseekmath,
    title        = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
    author       = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
    year         = 2024,
    eprint       = {arXiv:2402.03300},
}

Cite TRL as:

bibtex

@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}}
}