Reproduce “Generating Graph-Like Logical Rules for Knowledge Graph Reasoning via Diffusion Models” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Haoxiang Cheng, Yunfei Wang, Chao Chen, Kewei Cheng, Zhipeng Lin, Haoxuan Li, Changjun Fan, Shixuan Liu

GRiD, a framework from the National University of Defense Technology and Microsoft, introduces a method for generating graph-like logical rules for Knowledge Graph (KG) reasoning by leveraging discrete diffusion models. It successfully addresses the limited expressiveness of chain-like rules and the computational complexity of discovering richer rule structures, achieving competitive performance in Knowledge Graph Completion (KGC) tasks.

Key insights

  • Graph-like rules, by incorporating conjunctive constraints and complex topologies, capture distinct relational patterns that effectively complement chain-like rules, reducing ambiguity in KG reasoning.
  • Discrete diffusion models provide a viable generative paradigm to explore the vast rule space, overcoming the computational bottleneck of exhaustive search for complex rule structures.
  • Aligning generative processes with non-differentiable symbolic objectives requires a multi-stage optimization, with RL fine-tuning proving essential for generating high-quality, semantically meaningful rules.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.