RAMP_7B_mp_2_ratio_0.1

Model Details

RAMP (Raw-text Anchored Message Passing) recasts the LLM as a graph-native aggregation operator on text-rich graphs. This checkpoint is pretrained with 2 message passing layers and compression ratio 0.1.

• Base model: Qwen2.5-7B-Instruct
• Message passing layers: 2
• Compression ratio: 0.1

For training, evaluation, and usage details, see our GitHub repo.

Citation

@article{zhang2026llm,
  title={LLM as Graph Kernel: Rethinking Message Passing on Text-Rich Graphs},
  author={Zhang, Ying and Yu, Hang and Zhang, Haipeng and Di, Peng},
  journal={arXiv preprint arXiv:2603.14937},
  year={2026}
}