Reproduce “Mixture of Length and Pruning Experts for Knowledge Graphs Reasoning” 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
Enjun Du, Siyi Liu, Yongqi Zhang
MoKGR introduces a mixture-of-experts framework for Knowledge Graph reasoning, dynamically adjusting path exploration length and adaptively pruning paths based on structural and semantic relevance. The framework achieves superior accuracy and efficiency across diverse transductive and inductive benchmarks, including large-scale KGs like YAGO3-10.
Key insights
- Query-specific personalization of path exploration depth and breadth significantly enhances Knowledge Graph reasoning accuracy and computational efficiency.
- Effective path pruning requires a multi-faceted evaluation that integrates global prediction scores with both local structural patterns and semantic relevance to the query.
- A mixture-of-experts framework, with balanced expert utilization, allows for adaptive and complementary reasoning strategies that outperform single or fixed-strategy approaches.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.