G1 is the series of large language models trained on our benchmark Erdos for solving graph reasoning tasks, based on Qwen2.5-Instruct. We apply Group Relative Policy Optimization (GRPO) for reinforcement learning with supervised finetuning as a prelimary step.

G1 brings the following improvements:

Significant improvement on graph reasoning: G1 models achieve up to 46% improvement over baselines on Erdős, with the 7B variant matching OpenAI’s o3-mini and the 3B model surpassing Qwen2.5-72B-Instruct by notable margins.
Strong Generalization to unseen graph tasks: G1 exhibits zero-shot generalization on unseen graph tasks, improving performance on other graph reasoning benchmarks (GraphWiz, GraphArena) and real-world graphs (Cora, PubMed).
NO Compromise on general reasoning: Crucially, G1 preserves general reasoning ability (GSM8K, MATH, MMLU-Pro), proving its versatility.

This repo contains the G1-Direct-SFT-3B model, which has the following features:

Type: Causal Language Models
Training Stage: SFT
Architecture: the same with Qwen2.5-Instruct
Number of Parameters: 3.09B
Context Length: Full 32,768 tokens and generation 8192 tokens

For more details, please refer to our paper and GitHub.

Requirements

The model is trained based on Qwen/Qwen2.5-3B-Instruct. The code of Qwen2.5 has been in the latest Hugging face transformers and we advise you to use the latest version of transformers.

With transformers<4.37.0, you will encounter the following error:

KeyError: 'qwen2'

Quickstart

Here provides a code snippet with apply_chat_template to show you how to load the tokenizer and model and how to generate contents.

python

from transformers import AutoModelForCausalLM, AutoTokenizer

INSTRUCTION_TEMPLATE = """
    {instruction}

    Solve the above problem efficiently and clearly. The last line of your response should be of the following format: 'Therefore, the final answer is: $\\boxed{{ANSWER}}$. I hope it is correct' (without quotes) where ANSWER is just the final number or expression that solves the problem. Think step by step before answering.
    """.strip()

model_name = "PKU-ML/G1-Direct-SFT-3B"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "The task is to determine the degree centrality of a node in the graph.\n\n"\
        "Degree centrality for a node is the fraction of nodes it is connected to.\n\n"\
        "Here is an undirected graph containing nodes from 1 to 15. The edges are: (1, 15), (15, 11), (2, 3), (2, 6), (3, 6), (3, 7), (6, 7), (6, 8), (7, 8), (7, 14), (4, 10), (10, 5), (10, 12), (8, 14), (8, 9), (12, 11), (12, 13).\n\n"\
        "Question: What is the degree centrality of node 2 in the graph?\n\n"\
        "You need to format your answer as a float number."
messages = [
    {"role": "user", "content": INSTRUCTION_TEMPLATE.format(instruction=prompt)}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=4096,
    top_p=0.95,
    top_k=30,
    temperature=0.6
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

Evaluation & Performance

Detailed evaluation results are reported in this 📑 paper.

Citation

If you find our work helpful, feel free to give us a cite.

@article{guo2025g1,
  title={G1: Teaching LLMs to Reason on Graphs with Reinforcement Learning},
  author={Guo, Xiaojun and Li, Ang and Wang, Yifei and Jegelka, Stefanie and Wang, Yisen},
  journal={arXiv preprint arXiv:2505.18499},
  year={2025}
}