Reproduce “ReaRev: Adaptive Reasoning for Question Answering over Knowledge Graphs” 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
Costas Mavromatis, George Karypis
Researchers from the University of Minnesota introduce ReaRev, an adaptive reasoning approach for Knowledge Graph Question Answering. ReaRev employs BFS-emulating Graph Neural Networks and iterative instruction refinement to improve question answering performance on complex and incomplete knowledge graphs, showing an improvement of 4.1% in Hits@1 on Complex WebQuestions.
Key insights
- Allowing instructions to compete and dynamically determine their execution order based on knowledge graph structure improves reasoning flexibility.
- Iteratively refining graph traversal instructions using feedback from the knowledge graph during reasoning enhances robustness and corrects initial decoding errors.
- Combining Graph Neural Network-based traversal with adaptive instruction updating enables effective multi-hop reasoning on real-world, potentially incomplete, knowledge graphs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.