Qwen2.5 Bakeneko 32b Instruct V2

🚀 Qwen2.5 Bakeneko 32B Instruct V2 (rinna/qwen2.5-bakeneko-32b-instruct-v2)

This model is an instruction-tuned variant of rinna/qwen2.5-bakeneko-32b, fine-tuned with Chat Vector and Odds Ratio Preference Optimization (ORPO). It enhances instruction-following capabilities and demonstrates strong performance on Japanese benchmarks.

🚀 Quick Start

To quickly start using the Qwen2.5 Bakeneko 32B Instruct V2 model, you can refer to the following code example:

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "rinna/qwen2.5-bakeneko-32b-instruct-v2"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype=torch.bfloat16,
)

messages = [
    {"role": "system", "content": "あなたは誠実で優秀な日本人のアシスタントです。"},
    {"role": "user", "content": "ゲーム・小説・アニメに登場するアイテムボックスの特徴と、その原理を詳細に推測してください。"},
]
prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
input_ids = tokenizer.encode(
    prompt,
    add_special_tokens=False,
    return_tensors="pt"
).to(model.device)

outputs = model.generate(
    input_ids,
    max_new_tokens=1024,
    do_sample=True,
    temperature=0.7,
    top_k=20,
    top_p=0.8,
    repetition_penalty=1.05,
)

response = tokenizer.decode(outputs[0][input_ids.shape[-1]:], skip_special_tokens=True)
print(response)

✨ Features

• Enhanced Instruction-Following: Leveraging the advanced instruction-following capability from rinna/qwq-bakeneko-32b, this model enhances rinna/qwen2.5-bakeneko-32b-instruct without requiring additional reasoning processes.
• High Performance on Benchmarks: Demonstrates strong performance on Japanese benchmarks such as Japanese MT-Bench.
• Qwen2.5 Chat Format: Follows the Qwen2.5 chat format.

📚 Documentation

Model Overview

This model is an instruction-tuned variant of rinna/qwen2.5-bakeneko-32b, fine-tuned with Chat Vector and Odds Ratio Preference Optimization (ORPO).

Model Architecture

It is a 64-layer, 5120-hidden-size transformer-based language model. For a comprehensive understanding of the architecture, please refer to the Qwen2.5 Technical Report.

Training Process

The model was developed through a two-stage training process:

1. Model merging: The instruction-following capability was enhanced through a Chat Vector addition process.

   rinna/qwen2.5-bakeneko-32b-instruct + 0.8 * (rinna/qwq-bakeneko-32b - rinna/qwen2.5-bakeneko-32b)
   

   During this process, the embedding layer was excluded when performing the subtraction and addition of parameter vectors.
2. Distillation and ORPO: The merged model was further refined using ORPO, trained on 1.3k carefully curated samples generated by DeepSeek-R1.

Contributors

• Xinqi Chen
• Toshiaki Wakatsuki
• Kei Sawada

Release Date

February 19, 2025

📊 Benchmarking

For detailed benchmarking results, please refer to rinna's LM benchmark page (Sheet 20250319).

🔧 Technical Details

Tokenization

This model inherits the original Qwen/Qwen2.5-32B-Instruct tokenizer.

📖 References

@article{qwen2.5,
    title = {Qwen2.5 Technical Report}, 
    author = {An Yang and Baosong Yang and Beichen Zhang and Binyuan Hui and Bo Zheng and Bowen Yu and Chengyuan Li and Dayiheng Liu and Fei Huang and Haoran Wei and Huan Lin and Jian Yang and Jianhong Tu and Jianwei Zhang and Jianxin Yang and Jiaxi Yang and Jingren Zhou and Junyang Lin and Kai Dang and Keming Lu and Keqin Bao and Kexin Yang and Le Yu and Mei Li and Mingfeng Xue and Pei Zhang and Qin Zhu and Rui Men and Runji Lin and Tianhao Li and Tianyi Tang and Tingyu Xia and Xingzhang Ren and Xuancheng Ren and Yang Fan and Yang Su and Yichang Zhang and Yu Wan and Yuqiong Liu and Zeyu Cui and Zhenru Zhang and Zihan Qiu},
    journal = {arXiv preprint arXiv:2412.15115},
    year = {2024}
}

@misc{qwq32b,
    title = {QwQ-32B: Embracing the Power of Reinforcement Learning},
    url = {https://qwenlm.github.io/blog/qwq-32b/},
    author = {Qwen Team},
    month = {March},
    year = {2025}
}

@misc{deepseekai2025deepseekr1incentivizingreasoningcapability,
    title = {DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning}, 
    author = {DeepSeek-AI and Daya Guo and Dejian Yang and Haowei Zhang and Junxiao Song and Ruoyu Zhang and Runxin Xu and Qihao Zhu and Shirong Ma and Peiyi Wang and Xiao Bi and Xiaokang Zhang and Xingkai Yu and Yu Wu and Z. F. Wu and Zhibin Gou and Zhihong Shao and Zhuoshu Li and Ziyi Gao and Aixin Liu and Bing Xue and Bingxuan Wang and Bochao Wu and Bei Feng and Chengda Lu and Chenggang Zhao and Chengqi Deng and Chenyu Zhang and Chong Ruan and Damai Dai and Deli Chen and Dongjie Ji and Erhang Li and Fangyun Lin and Fucong Dai and Fuli Luo and Guangbo Hao and Guanting Chen and Guowei Li and H. Zhang and Han Bao and Hanwei Xu and Haocheng Wang and Honghui Ding and Huajian Xin and Huazuo Gao and Hui Qu and Hui Li and Jianzhong Guo and Jiashi Li and Jiawei Wang and Jingchang Chen and Jingyang Yuan and Junjie Qiu and Junlong Li and J. L. Cai and Jiaqi Ni and Jian Liang and Jin Chen and Kai Dong and Kai Hu and Kaige Gao and Kang Guan and Kexin Huang and Kuai Yu and Lean Wang and Lecong Zhang and Liang Zhao and Litong Wang and Liyue Zhang and Lei Xu and Leyi Xia and Mingchuan Zhang and Minghua Zhang and Minghui Tang and Meng Li and Miaojun Wang and Mingming Li and Ning Tian and Panpan Huang and Peng Zhang and Qiancheng Wang and Qinyu Chen and Qiushi Du and Ruiqi Ge and Ruisong Zhang and Ruizhe Pan and Runji Wang and R. J. Chen and R. L. Jin and Ruyi Chen and Shanghao Lu and Shangyan Zhou and Shanhuang Chen and Shengfeng Ye and Shiyu Wang and Shuiping Yu and Shunfeng Zhou and Shuting Pan and S. S. Li and Shuang Zhou and Shaoqing Wu and Shengfeng Ye and Tao Yun and Tian Pei and Tianyu Sun and T. Wang and Wangding Zeng and Wanjia Zhao and Wen Liu and Wenfeng Liang and Wenjun Gao and Wenqin Yu and Wentao Zhang and W. L. Xiao and Wei An and Xiaodong Liu and Xiaohan Wang and Xiaokang Chen and Xiaotao Nie and Xin Cheng and Xin Liu and Xin Xie and Xingchao Liu and Xinyu Yang and Xinyuan Li and Xuecheng Su and Xuheng Lin and X. Q. Li and Xiangyue Jin and Xiaojin Shen and Xiaosha Chen and Xiaowen Sun and Xiaoxiang Wang and Xinnan Song and Xinyi Zhou and Xianzu Wang and Xinxia Shan and Y. K. Li and Y. Q. Wang and Y. X. Wei and Yang Zhang and Yanhong Xu and Yao Li and Yao Zhao and Yaofeng Sun and Yaohui Wang and Yi Yu and Yichao Zhang and Yifan Shi and Yiliang Xiong and Ying He and Yishi Piao and Yisong Wang and Yixuan Tan and Yiyang Ma and Yiyuan Liu and Yongqiang Guo and Yuan Ou and Yuduan Wang and Yue Gong and Yuheng Zou and Yujia He and Yunfan Xiong and Yuxiang Luo and Yuxiang You and Yuxuan Liu and Yuyang Zhou and Y. X. Zhu and Yanhong Xu and Yanping Huang and Yaohui Li and Yi Zheng and Yuchen Zhu and Yunxian Ma and Ying Tang and Yukun Zha and Yuting Yan and Z. Z. Ren and Zehui Ren and Zhangli Sha and Zhe Fu and Zhean Xu and Zhenda Xie and Zhengyan Zhang and Zhewen Hao and Zhicheng Ma and Zhigang Yan and Zhiyu Wu and Zihui Gu and Zijia Zhu and Zijun Liu and Zilin Li and Ziwei Xie and Ziyang Song and Zizheng Pan and Zhen Huang and Zhipeng Xu and Zhongyu Zhang and Zhen Zhang},
    year = {2025},
    eprint = {2501.12948},
    archivePrefix = {arXiv},
    primaryClass = {cs.CL},
    url = {https://arxiv.org/abs/2501.12948}, 
}

@article{huang2023chat,
    title = {Chat Vector: A Simple Approach to Equip LLMs with Instruction Following and Model Alignment in New Languages},
    author = {Huang, Shih-Cheng and Li, Pin-Zu and Hsu, Yu-Chi and Chen, Kuang-Ming and Lin, Yu Tung and Hsiao, Shih-Kai and Tzong-Han Tsai, Richard and Lee, Hung-yi},
    year = {2023},
    url = {https://arxiv.org/abs/2310.04799}
}

@inproceedings{hong2024orpo,
    title = {ORPO: Monolithic Preference Optimization without Reference Model},
    author = {Hong, Jiwoo and Lee, Noah and Thorne, James},
    booktitle = {Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing},
    pages = {11170--11189},
    year = {2024}
}

📄 License

The Apache License, Version 2.0