metadata

task_categories:
  - graph-ml
license: mit
tags:
  - multi-agent-systems
  - benchmark
  - llms
dataset_info:
  features:
    - name: 'Unnamed: 0'
      dtype: int64
    - name: graph_generator
      dtype: string
    - name: num_nodes
      dtype: int64
    - name: index
      dtype: int64
    - name: graph
      dtype: string
  splits:
    - name: train
      num_bytes: 517975
      num_examples: 117
  download_size: 127270
  dataset_size: 517975
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

AgentsNet

This repository contains the graph instances used in the AgentsNet: Coordination and Collaborative Reasoning in Multi-Agent LLMs paper.

AgentsNet is a new benchmark for multi-agent reasoning, designed to measure the ability of multi-agent systems to collaboratively form strategies for problem-solving, self-organization, and effective communication given a network topology. It draws inspiration from classical problems in distributed systems and graph theory.

Paper: AgentsNet: Coordination and Collaborative Reasoning in Multi-Agent LLMs
Project Page: https://agentsnet.graphben.ch
GitHub Repository: https://github.com/ashleve/lightning-hydra-template

Dataset

The dataset consists of synthetic graphs generated using various random graph models. It serves as the input for all experiments in the benchmark.

Citation

If you use this dataset, please cite the associated paper:

@misc{grötschla2025agentsnetcoordinationcollaborativereasoning,
      title={AgentsNet: Coordination and Collaborative Reasoning in Multi-Agent LLMs}, 
      author={Florian Grötschla and Luis Müller and Jan Tönshoff and Mikhail Galkin and Bryan Perozzi},
      year={2025},
      eprint={2507.08616},
      archivePrefix={arXiv},
      primaryClass={cs.MA},
      url={https://arxiv.org/abs/2507.08616}, 
}