Researchers have developed Gracr, a novel compiler that translates graded modal logic formulas into Graph Neural Network (GNN) weights. This approach replaces traditional GNN training with compilation, ensuring that the model's behavior precisely replicates the logic formulas. To rigorously evaluate GNN explainers, a new benchmark called GracrBench has been introduced, which uses these compiled GNNs to provide an exact ground truth for explanations. Experiments using GracrBench have revealed that many existing explainers lack robustness against indirect influences or alternative implementations of the same logic. AI
IMPACT Introduces a more rigorous evaluation framework for GNN explainability, potentially leading to more reliable and robust AI systems.
RANK_REASON Academic paper introducing a new method and benchmark for evaluating graph neural network explainability. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Gotit.pub
- Gracr
- GracrBench
- graph neural networks
- Hugging Face
- IArxiv
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
