README.md

metadata

license: mit
language:
  - en
size_categories:
  - n<1K
pretty_name: Adaptive Graph Pruning Training Dataset
dataset_info:
  - config_name: coding
    features:
      - name: task
        dtype: string
      - name: edge_weight
        sequence:
          sequence: float32
      - name: mask
        sequence: float32
    splits:
      - name: train
        num_bytes: 85430
        num_examples: 140
  - config_name: general_reasoning
    features:
      - name: task
        dtype: string
      - name: edge_weight
        sequence:
          sequence: float32
      - name: mask
        sequence: float32
    splits:
      - name: train
        num_bytes: 97455
        num_examples: 138
  - config_name: math_reasoning
    features:
      - name: task
        dtype: string
      - name: edge_weight
        sequence:
          sequence: float32
      - name: mask
        sequence: float32
    splits:
      - name: train
        num_bytes: 31794
        num_examples: 94
configs:
  - config_name: coding
    data_files:
      - split: train
        path: coding/train.json
  - config_name: general_reasoning
    data_files:
      - split: train
        path: general_reasoning/train.json
  - config_name: math_reasoning
    data_files:
      - split: train
        path: math_reasoning/train.json

Dataset Card for AGP-Training

Dataset Description

• Homepage: https://resurgamm.github.io/AGP/
• Repository: https://github.com/Resurgamm/AGP
• Paper: https://arxiv.org/abs/2506.02951

Dataset Summary

AGP-Training is the dataset for training Adaptive Graph Pruning (AGP). We open-source our training set so that you can reproduce our work.

For further usage, please refer to our repository

Citation

If you find AGP helpful in your research, please consider citing:

@article{li2025adaptive,
  title={Adaptive Graph Pruning for Multi-Agent Communication},
  author={Li, Boyi and Zhao, Zhonghan and Lee, Der-Horng and Wang, Gaoang},
  journal={arXiv preprint arXiv:2506.02951},
  year={2025}
}