LLM4WM: Adapting LLM for Wireless Multi-Tasking

Liu, Xuanyu, et al. "LLM4WM: Adapting LLM for Wireless Multi-Tasking." IEEE Transactions on Machine Learning in Communications and Networking (2025). [paper]

📰 News

• SoM Challenge 2025 — WiFo-empowered Wireless Multi-tasking announced!
  Our PCNI Lab, Peking University is co-hosting the SoM Challenge 2025 with ITU (sponsored by Huawei), focusing on WiFo-empowered wireless multi-task learning at the wireless physical layer.

  Key Dates (UTC+8)

  • 🏁 Challenge launch: Sep 30, 2025
  • 📥 Registration deadline: Dec 15, 2025, 23:59
  • 🛠 Development phase deadline: Feb 1, 2026, 23:59

  👉 Official announcement & details: WiFo Discussion #8

  We warmly welcome everyone to register, follow the challenge, and ask questions in the discussion forum 🎉

Dependencies and Installation

• Python 3.8 (Recommend to use Anaconda)
• Pytorch 2.0.0
• NVIDIA GPU + CUDA
• Python packages: pip install -r requirements.txt

Dataset Preparation

The test datasets used in this paper is generated by QuaDRiGa, and it can be downloaded in the following links. Testing Dataset Link: [Baidu Cloud] | [Hugging Face] Training Dataset Link: [Baidu Cloud] | [Hugging Face]

Get Started

Step 1: Prepare the Files

• Dataset: Download the dataset and place it under the dataset/ folder in the root directory.
• GPT-2 Weights: Download the GPT-2 weights and put them into the pretrain/gpt2/ folder.
• LLM4WM Weights: Download our provided pretrained weights of LLM4WM( Baidu Cloud | Hugging Face) and store them in the Weights/ folder.

Step 2: Run Inference

Once all the required files are in place, you can evaluate our pretrained model with:

python inference.py

Citation

If you find this repo helpful, please cite our paper.

@article{liu2025llm4wm,
  title={LLM4WM: Adapting LLM for Wireless Multi-Tasking},
  author={Liu, Xuanyu and Gao, Shijian and Liu, Boxun and Cheng, Xiang and Yang, Liuqing},
  journal={IEEE Transactions on Machine Learning in Communications and Networking},
  year={2025}
}