Re-explanation context perturbation causes self-inconsistency in SI-GNN explanations via latent signal assignment; Self-Denoising calibrates them with one extra forward pass and improves quality per experiments on standard benchmarks.
Graph neural networks including sparse interpretability.ArXiv
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Why Self-Inconsistency Arises in GNN Explanations and How to Exploit It
Re-explanation context perturbation causes self-inconsistency in SI-GNN explanations via latent signal assignment; Self-Denoising calibrates them with one extra forward pass and improves quality per experiments on standard benchmarks.