GNN Explainability

Temporal graph neural networks (TGNNs) have gained significant traction for solving real-world temporal graph tasks. However, their interpretability remains limited, as most TGNNs fail to identify which historical interactions most influence a given prediction. Despite promising progress on interpretable TGNNs, existing methods predominantly focus on previously seen historical interactions, which we term stability patterns, while overlooking newly emerging first-time interactions, which we term transition patterns. Both types of patterns are essential for faithful temporal explanations. To address this limitation, we propose ST-TGExplainer, a self-explainable TGNN that disentangles Stability and Transition patterns in temporal graphs for a more faithful Temporal GNN Explainer. Guided by a disentangled information bottleneck objective, ST-TGExplainer learns a compact explanatory subgraph that remains predictive of the event label while explicitly suppressing label-conditioned redundancy between stability and transition patterns. Extensive experiments demonstrate that ST-TGExplainer achieves strong predictive performance and yields more faithful explanations. Code is available at https://github.com/hjchen-hdu/ST-TGExplainer.

We introduce B-cos GNNs, an inherently explainable class of graph neural networks whose predictions decompose exactly into per-node, per-feature contributions via a single input-dependent linear map. B-cos GNNs use linear (sum-based) aggregation and replace non-linear message and update functions with B-cos transforms. This induces meaningful, task-specific weight-input alignment that is directly accessible through the model's dynamic linearity. Instance-level explanations follow from a single forward and backward pass, requiring no auxiliary explainer, modified learning objective, or perturbation procedure. Instantiated as a GIN, our approach trades small losses in predictive accuracy for state-of-the-art explainability across diverse synthetic and real-world benchmarks, producing explanations orders of magnitude faster than post-hoc baselines.

Node representation learning, such as Graph Neural Networks (GNNs), has emerged as a pivotal method in machine learning. The demand for reliable explanation generation surges, yet unsupervised models remain underexplored. To bridge this gap, we introduce a method for generating counterfactual (CF) explanations in unsupervised node representation learning. We identify the most important subgraphs that cause a significant change in the k-nearest neighbors of a node of interest in the learned embedding space upon perturbation. The k-nearest neighbor-based CF explanation method provides simple, yet pivotal, information for understanding unsupervised downstream tasks, such as top-k link prediction and clustering. Consequently, we introduce UNR-Explainer for generating expressive CF explanations for Unsupervised Node Representation learning methods based on a Monte Carlo Tree Search (MCTS). The proposed method demonstrates superior performance on diverse datasets for unsupervised GraphSAGE and DGI.

Graph Neural Networks (GNNs) have advanced significantly in handling graph-structured data, but a comprehensive framework for evaluating explainability remains lacking. Existing evaluation frameworks primarily involve post-hoc explanations, and operate in the setting where multiple methods generate a suite of explanations for a single model. This makes comparison of explanations across models difficult. Evaluation of inherently interpretable models often targets a specific aspect of interpretability relevant to the model, but remains underdeveloped in terms of generating insight across a suite of measures. We introduce AIM, a comprehensive framework that addresses these limitations by measuring Accuracy, Instance-level explanations, and Model-level explanations. AIM is formulated with minimal constraints to enhance flexibility and facilitate broad applicability. Here, we use AIM in a pipeline, extracting explanations from inherently interpretable GNNs such as graph kernel networks (GKNs) and prototype networks (PNs), evaluating these explanations with AIM, identifying their limitations and obtaining insights to their characteristics. Taking GKNs as a case study, we show how the insights obtained from AIM can be used to develop an updated model, xGKN, that maintains high accuracy while demonstrating improved explainability. Our approach aims to advance the field of Explainable AI (XAI) for GNNs, providing more robust and practical solutions for understanding and improving complex models.

We propose a subgraph importance estimation method for pretrained Graph Neural Networks (GNNs) on graph-level tasks, formulated as a linear Group Lasso regression problem in the embedding space. Our method effectively leverages prior domain knowledge of graph substructures, while remaining independent of the specific form of the output layer or readout function used in the GNN architecture, and it does not require access to ground-truth target labels. Experiments on real-world graph datasets demonstrate that our method consistently outperforms existing baselines in subgraph importance estimation. Furthermore, we extend our method to identify important nodes within the graph.

Graph Neural Networks (GNNs) excel at relational reasoning but face two persistent challenges: the lack of interpretable attribution for heterogeneous node types, and the computational overhead of message passing over large, noisy graphs. We propose the Hierarchical Attention-based Heterogeneous GNN (HA-HeteroGNN), a framework that addresses both issues through a unied explainability-to-pruning pipeline. A two-tier attention mechanism separates sensor-level and context-level computation across 16 node types and 18 edge types, producing per-node relevance scores via an attention-based GNN Explainer without requiring gradient backpropagation. These relevance scores then serve as a principled pruning criterion: removing nodes identied as consistently uninformative yields a 27% reduction in graph edges while simultaneously improving classication accuracy by 2.46.1% across all model variants, challenging the conventional assumption that pruning necessarily trades accuracy for eciency. Experiments on a 50,000-record synthetic dataset spanning 11 report categories demonstrate 97.5% cross-strategy explanation stability and domain consistent sensor attribution, with training-time reductions of up to 43.9% and real-time inference latency of approximately 5860 ms per sample.

Protein language models such as ESM-2 learn rich residue representations that achieve strong performance on protein function prediction, but their features remain difficult to interpret as structural $\&$ evolutionary signals are encoded in dense latent spaces. We propose a plug-$\&$-play framework that projects ESM-2 representations onto protein contact graphs $\&$ applies $\textbf{SoftBlobGIN}$, a lightweight Graph Isomorphism Network with differentiable Gumbel-softmax substructure pooling, to perform structure-aware message passing $\&$ learn coarse functional substructures for downstream prediction tasks. Across enzyme classification, SoftBlobGIN achieves 92.8% accuracy $\&$ 0.898 macro-F1. Unlike post hoc analysis of protein language models alone, our method produces directly auditable structural explanations: GNNExplainer recovers biologically meaningful active-site residues, spatially localized functional clusters, $\&$ catalytic contact patterns. On binding-site detection, SoftBlobGIN improves residue AUROC from $0.885$ using an ESM-2 linear probe to $0.983$, indicating that these structural explanations are not recoverable from language-model features alone. Learned blob partitions provide an additional layer of interpretability by automatically grouping residues into functional substructures, with blobs containing annotated active-site residues showing $1.85\times$ higher importance than other blobs ($ρ{=}0.339$, $p{=}0.009$), without any active-site supervision. Our framework requires no retraining of the language model, adds only $\sim$1.1M parameters, $\&$ generalises across ProteinShake tasks, achieving $F_{\max}$ of $0.733$ on Gene Ontology prediction $\&$ AUROC of $0.969$ on binding-site detection. We position this as an interpretable structural companion to protein language models that makes their predictions more transparent $\&$ auditable.

Graph neural networks (GNNs) are increasingly used to model biological systems, yet the reliability of post-hoc explanation methods for recovering meaningful molecular mechanisms remains unclear. Here, we systematically evaluate four widely used approaches: Saliency Attribution (SA), Integrated Gradients (IG), GNNExplainer, and Layer-wise Relevance Propagation (LRP) for identifying disease-relevant structure in breast cancer RNA-seq data projected onto a protein-protein interaction network. Using synthetic benchmarks with known ground-truth motifs, we show that explanation methods recover distinct signal organizations: SA performs best for sparse single-node drivers, whereas IG and LRP preferentially recover distributed pathway-like and cascade-like signals. In TCGA BRCA data, we identify a consistent topological signature of disease-associated hubs in which attribution peaks in the immediate 1-hop neighborhood and decays across successive network shells, a pattern most pronounced for IG and LRP and associated with strong enrichment of known cancer hubs. We further observe a trade-off between local hub enrichment and global gene ranking performance, with IG optimizing local enrichment and SA achieving superior global discrimination. Motivated by these complementary behaviors, we introduce a framework combining a shell-based hub score with consensus ranking across explainers. Consensus scores improve prioritization of canonical cancer genes (TP53, BRCA1, ESR1, MYC), reduce dependence on node degree, and, especially when tuned, outperform individual methods. Pathway enrichment further reveals improved recovery of biologically coherent cancer programs, including ERBB2, RTK, MAPK, immune, and cytokine signaling. Together, these results demonstrate that topology-aware integration of graph explanations can improve biological interpretability and biologically relevant molecular recovery.

Recent work has observed that explanations produced by Self-Interpretable Graph Neural Networks (SI-GNNs) can be self-inconsistent: when the model is reapplied to its own explanatory graph subset, it may produce a different explanation. However, why self-inconsistency arises remains poorly understood. In this work, we first identify re-explanation-induced context perturbation as the direct cause of score variation. We then introduce a latent signal assignment hypothesis to explain why only some edges are sensitive to this perturbation, and analyze how conciseness regularization affects latent signal assignment. Given that self-inconsistent edges do not provide stable evidence for the model's prediction, we propose Self-Denoising (SD), a model-agnostic and training-free post-processing strategy that calibrates explanations with only one additional forward pass. Experiments across representative SI-GNN frameworks, backbone architectures, and benchmark datasets support our hypothesis and show that SD consistently improves explanation quality while adding only about 4--6% computational overhead in practice.

Regulatory frameworks such as GDPR increasingly require that ML predictions be accompanied by post-hoc explanations, even when raw data and trained models cannot be released. Differential privacy (DP) is the standard mitigation for the residual privacy risk of releasing these explanations. We show that DP is not sufficient: an adversary observing only DP-perturbed GNN explanations can reconstruct hidden graph structure with high accuracy. Our attack, PRIVX, exploits the fact that the Gaussian DP mechanism is a single DDPM forward step at known noise level σ(ε), recasting reconstruction as reverse diffusion conditioned on the corrupted signal, a principled Bayesian denoiser under known DP corruption. We formalise a stratified adversary model parameterised by (M, \hatε, \hatδ, S, ρ) that interpolates between oblivious and oracle attackers, and derive endpoint-matched two-sided bounds on reconstruction AUC. For practitioners, we provide regime-stratified guidance on explainer choice: on homophilic graphs, neighbourhood-aggregating explainers (GraphLIME, GNNExplainer) leak more structure than per-node gradient explainers under the same DP budget; on strongly heterophilic graphs the ordering reverses. We introduce PRIVF as an auxiliary diagnostic sharing the same diffusion backbone to decompose leakage into explainer-induced and intrinsic graph-distribution components. Experiments across seven benchmarks, three DP mechanisms, and three GNN backbones show PRIVX achieves AUC above 0.7 at ε = 5 on five of seven datasets, with the attack succeeding well within typically deployed privacy budgets.

Graph Neural Networks (GNNs) achieve strong performance on node classification tasks but remain difficult to interpret, particularly with respect to which input features drive their predictions. Existing global GNN explainers operate at the structural level identifying recurring subgraph motifs, but none explain model behaviour globally at the level of input node attributes. We propose GRAFT, a posthoc global explanation framework that identifies class-level feature importance profiles for GNNs. The method combines diversity-guided exemplar selection, Integrated Gradients-based attribution, and aggregation to construct a global view of feature influence for each class, which can be further expressed as concise natural language rules using a large language model with self-refinement. We evaluate GRAFT across multiple datasets, architectures, and experimental settings, demonstrating its effectiveness in capturing model-relevant features, supporting bias analysis, and enabling feature-efficient transfer learning. In addition, we introduce a structured human evaluation protocol to assess the interpretability of generated rules along dimensions such as accuracy and usefulness. Our results suggest that GRAFT provides a practical and interpretable approach for analysing feature-level behaviour in GNNs, bridging quantitative attribution with human-understandable explanations.

Graph neural networks such as ParticleNet and transformer based networks on point clouds such as ParticleTransformer achieve state-of-the-art performance on jet tagging benchmarks at the Large Hadron Collider, yet the physical reasoning behind their predictions remains opaque. We present different methods, i.e. perturbation-based (GNNExplainer), Shapley-value-based (GNNShap), and gradient-based (GRADCam); adapted to operate on LundNet's Lund-plane graph representation. Leveraging the fact that each node in the Lund plane corresponds to a physically meaningful parton splitting, we construct Monte Carlo truth explanation masks and introduce a physics-informed evaluation framework that goes beyond standard fidelity metrics. We perform the analysis in three transverse-momentum bins ($\mathrm{p_T} \in [500,700]$, $[800,1000]$, and the inclusive region $[500,1000]$ GeV), revealing how explanation quality and focus shift between non-perturbative and perturbative regimes. We further quantify the correlation between explainer-assigned node importance and classical jet substructure observables -- $N$-subjettiness ratios $τ_{21}$ and $τ_{32}$ and the energy correlation functions -- establishing the degree to which the model has learned known QCD features. We find that overall the weight assigned by explainability methods has a correlation with analytic observables, with expected shift across different phase space regimes, indicating that a trained neural network indeed learns some aspects of jet-substructure moments. Our open-source implementation enables reproducible explainability studies for graph-based jet taggers.

Temporal Graph Neural Networks (TGNNs) have become increasingly popular in recent years due to their superior predictive performance by combining both spatial and temporal information. However, how these models utilize the information to make predictions is rather unexplored, leading to potentially faulty or biased models. This work introduces two novel model-agnostic explainers for local explanations of TGNNs based on Shapley and Owen values. The first method, an event-level (edge-level) Shapley explainer, applies the KernelSHAP algorithm to estimate contribution scores for individual temporal events, providing interpretable descriptions for model behavior. The second, a feature-level Shapley explainer, extends this framework by decomposing event-level Shapley values into Owen values, and thereby uncovers hierarchical dependencies of the event and its features. The explainers outperform SOTA explainers on different metrics and datasets. Additionally, the Feature Explainer reveals a faulty extraction of actual timestamps of a commonly used TGAT implementation, helping to further understand performance drops on very sparse explanations.

Graph neural networks achieve strong node-classification accuracy, but learned message passing entangles ego attributes, neighborhood smoothing, high-pass graph differences, class geometry, and classifier-boundary effects inside opaque representations. This obscures why nodes are classified as they are and which graph-learning mechanisms a dataset requires. We propose WG-SRC, a white-box signal-subspace probe for prediction and graph dataset diagnosis. WG-SRC replaces learned message passing with a fixed, named graph-signal dictionary containing raw features, row- and symmetric-normalized low-pass propagation, and high-pass graph differences. It combines Fisher coordinate selection, class-wise PCA subspaces, closed-form multi-alpha ridge classification, and validation-based score fusion, so prediction and analysis rely on explicit class subspaces, energy-controlled dimensions, and closed-form linear decisions. As a white-box graph-learning instrument, WG-SRC uses predictive performance to validate its diagnostics. Across six node-classification datasets, it remains competitive with reproduced baselines and achieves positive average gain under aligned splits. Its atlas decomposes behavior into raw-feature, low-pass, high-pass, class-geometric, and ridge-boundary components. The resulting fingerprints distinguish low-pass-dominated Amazon graphs, mixed high-pass and class-geometrically complex Chameleon behavior, and raw- or boundary-sensitive WebKB graphs. Aligned interventions show when high-pass blocks act as removable noise, when raw or graph-derived signals should be preserved, and when ridge correction matters. WG-SRC therefore serves both as a functioning white-box classifier and as a dataset-fingerprinting probe, enabling fingerprint-conditioned analysis of how black-box model components behave under different graph-signal conditions.

Artificial Intelligence (AI) weather models are improving rapidly, and their forecasts are already competitive with long-established traditional Numerical Weather Prediction (NWP). To build confidence in this new methodology, it is critical that we understand how these predictions are generated. This is a huge challenge as these AI weather models remain largely black boxes. In other areas of Machine Learning (ML), mechanistic interpretability has emerged as a framework for understanding ML predictions by analysing the building blocks responsible for them. Here we present an open-source, highly adaptable tool which incorporates concepts from mechanistic interpretability. The tool organises internal latent representations from the model processor and allows for initial analyses, including cosine similarity and Principal Component Analysis (PCA), enabling the user to identify directions in latent space potentially associated with meteorological features. Applying our tool to the graph neural network GraphCast, we present preliminary case studies for mid-latitude synoptic-scale waves and specific humidity. These demonstrate the tool's ability to identify linear combinations of latent channels that appear to correspond to interpretable features.

The trust in the predictions of Graph Neural Networks is limited by their opaque reasoning process. Prior methods have tried to explain graph networks via concept-based explanations extracted from the latent representations obtained after message passing. However, these explanations fall short of explaining the message passing process itself. To this aim, we propose the Concept Graph Convolution, the first graph convolution designed to operate on node-level concepts for improved interpretability. The proposed convolutional layer performs message passing on a combination of raw and concept representations using structural and attention-based edge weights. We also propose a pure variant of the convolution, only operating in the concept space. Our results show that the Concept Graph Convolution allows to obtain competitive task accuracy, while enabling an increased insight into the evolution of concepts across convolutional steps.

Graph representation learning has achieved notable success in encoding graph-structured data into latent vector spaces, enabling a wide range of downstream tasks. However, these node representations remain opaque and difficult to interpret. Existing explainability methods primarily focus on supervised settings or on explaining individual representation dimensions, leaving a critical gap in explaining the overall structure of node representations. In this paper, we propose TACENR (Task-Agnostic Contrastive Explanations for Node Representations), a local explanation method that identifies not only attribute features but also proximity and structural ones that contribute the most in the representation space. TACENR builds on contrastive learning, through which we learn a similarity function in the representation space, revealing which are the features that play an important role in the representation of a node. While our focus is on task-agnostic explanations, TACENR can be applied to supervised scenarios as well. Experimental results demonstrate that proximity and structural features play a significant role in shaping node representations and that our supervised variant performs comparably to existing task-specific approaches in identifying the most impactful features.

The reasoning process of Graph Neural Networks is complex and considered opaque, limiting trust in their predictions. To alleviate this issue, prior work has proposed concept-based explanations, extracted from clusters in the model's node embeddings. However, a limitation of concept-based explanations is that they only explain the node embedding space and are obscured by pooling in graph classification. To mitigate this issue and provide a deeper level of understanding, we propose the Subgraph Concept Network. The Subgraph Concept Network is the first graph neural network architecture that distils subgraph and graph-level concepts. It achieves this by performing soft clustering on node concept embeddings to derive subgraph and graph-level concepts. Our results show that the Subgraph Concept Network allows to obtain competitive model accuracy, while discovering meaningful concepts at different levels of the network.

Jet identification plays a central role in analyzing data from high-energy collider experiments. While deep learning has improved jet classification, it often lacks interpretability. We introduce the Kinematic Interaction Graph Network (KIGNet), a graph neural network that integrates kinematic variables into jet classification by constructing four graph representations per jet, each weighted by a distinct variable: angular separation ($Δ$), relative transverse momentum ($k_T$), momentum fraction ($z$), and invariant mass squared ($m^2$). Three of these ($Δ$, $k_T$, $z$) are motivated by the Lund jet plane, grounded in perturbative QCD factorization; the fourth ($m^2$) adds complementary mass-scale sensitivity for heavy-flavor identification. Using Gradient-weighted Class Activation Mapping (Grad-CAM), we determine which variables dominate classification. Angular separation and relative transverse momentum account for about 76% of the total Grad-CAM attribution (40.72% and 35.67%), with momentum fraction and invariant mass contributing the remaining 24%. This hierarchy is consistent with the soft-collinear structure of QCD radiation in the training data, showing that the network learns physically interpretable representations rather than spurious correlations. On the JetClass dataset, KIGNet achieves a macro-accuracy of 95.07%, macro-AUC of 96.61%, and macro-AUPR of 81.52%, relative improvements of 2.45%, 3.40%, and 19.11% over the state-of-the-art baseline. On the Aspen Open Jets dataset of real CMS collision data, KIGNet produces substantially more structured latent representations than the baseline, reducing the Davies-Bouldin Index by 52.15% ($0.8395 \rightarrow 0.4017$) and increasing the Dunn Index by 42.33% ($0.0189 \rightarrow 0.0269$), confirming that physics-informed kinematic encoding generalizes beyond idealized simulation to experimental detector conditions.

Explainable artificial intelligence approaches accelerate drug discovery by improving molecular representation learning, identifying key molecular structures, and rationalizing drug property prediction. However, developing end-to-end explainable models for target-specific structure-activity relationship modeling remains challenging because compound-protein interaction data are often limited for individual targets, and small changes in chemical substituents or local structural motifs can cause large differences in molecular properties. Therefore, effectively leveraging structural and property information to identify key moieties associated with compound-protein affinity is essential. We propose a graph neural network (GNN) framework that uses property and structural information from activity-cliff molecule pairs targeting specific proteins to predict compound-protein affinity, measured by half-maximal inhibitory concentration (IC50), and explain property differences. To improve explainability, we trained GNNs with structure-aware loss functions using group lasso and sparse group lasso regularization, which prune and highlight molecular subgraphs relevant to activity differences. We applied this framework to activity-cliff data from molecules targeting six tyrosine-protein kinases across the Src, Abl, and Tec families, as well as anaplastic lymphoma kinase. Integrating common- and uncommon-node information with sparse group lasso improved target-specific molecular property prediction, producing lower root mean square errors and higher Pearson correlation coefficients. Regularization also enhanced GNN feature attribution by improving graph-level global direction scores and atom-level coloring accuracy. These results support more interpretable drug discovery pipelines, particularly for identifying critical molecular substructures during lead optimization.