Research map / Graph learning / Graph self-supervised learning
D4Explainer: In-distribution Explanations of Graph Neural Network via Discrete Denoising Diffusion
Jialin Chen, Shirley Wu, Abhijit Gupta, Zhitao Ying
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| reduce reverse sampling steps T from 50 to 10 (K=20) (ablation) | D4Explainer · Mutag model-level | target class probability: 0.812 -> 0.524 | Hurt |
| increase T from 50 to 100 (K=20) (ablation) | D4Explainer · Mutag model-level | target class probability: 0.812 -> 0.832 | Improved |
| reduce candidates K from 20 to 10 (T=50) (ablation) | D4Explainer · Mutag model-level | target class probability: 0.812 -> 0.799 | Hurt |
| increase K from 20 to 30 (T=50) (ablation) | D4Explainer · Mutag model-level | target class probability: 0.812 -> 0.823 | Improved |
| vary K and T on Tree-Cycle (ablation) | D4Explainer · Tree-Cycle model-level | target class probability: 0.987 -> 0.997 | No clear effect |
| D4Explainer vs CF-GNNExplainer on counterfactual accuracy | D4Explainer · Cornell | CF-ACC AUC: 0.328 -> 0.623 | Improved |
| D4Explainer vs CF-GNNExplainer on BA-3Motif | D4Explainer · BA-3Motif | CF-ACC AUC: 0.302 -> 0.912 | Improved |
| D4Explainer vs CF-GNNExplainer on NCI1 (only small gap) | D4Explainer · NCI1 | CF-ACC AUC: 0.715 -> 0.737 | Improved |
| D4Explainer vs CF-GNNExplainer on Mutag | D4Explainer · Mutag | CF-ACC AUC: 0.797 -> 0.765 | No clear effect |
| D4Explainer vs CF-GNNExplainer MMD sum | D4Explainer · BBBP | MMD Sum: 0.1264 -> 0.0861 | Improved |
| D4Explainer vs XGNN sparsity | D4Explainer · Mutag model-level | density: 0.537 -> 0.278 | Improved |
| D4Explainer vs XGNN with 5 nodes on Tree-Cycle | D4Explainer · Tree-Cycle model-level | target class probability: 0.968 -> 0.991 | No clear effect |
Later papers that compared with this method
2 later accepted papers report D4Explainer as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Generating In-Distribution Counterfactual Explanation for Graph Neural Networks (AAAI 2026 · better on 15 of 15 shared results)
- Paradigm Shift of GNN Explainer from Label Space to Prototypical Representation Space (ICLR 2026 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Generating In-Distribution Counterfactual Explanation for Graph Neural Networks (AAAI 2026 · similarity 0.92)
- Towards Robust Fidelity for Evaluating Explainability of Graph Neural Networks (ICLR 2024 · similarity 0.90)
- Generating In-Distribution Proxy Graphs for Explaining Graph Neural Networks (ICML 2024 · similarity 0.89)
- RegExplainer: Generating Explanations for Graph Neural Networks in Regression Tasks (NeurIPS 2024 · similarity 0.89)
- Self-Interpretable Graph Learning with Sufficient and Necessary Explanations (AAAI 2024 · similarity 0.89)
- Stratified GNN Explanations through Sufficient Expansion (AAAI 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
