Layer-Wise Relevance Propagation

Also known as LRP

Layer-wise relevance propagation studies how to explain a neural network’s prediction by passing relevance scores backward through its layers to the inputs. Researchers design propagation rules for convolutional networks, transformers with attention and residual connections, graph neural networks, and models with specialized geometries, then assess explanations using faithfulness, localization, and domain-specific checks. Applications include vision, language, biology, neuroscience, and scientific modeling, where explanations can help reveal model failures or test whether decisions reflect meaningful evidence. A central challenge is making attributions both efficient and reliable across architectures: relevance conservation alone does not guarantee faithful explanations, and results can depend strongly on the model and evaluation method.

Latest papers

16 papers1 paper in the last four weeks, against 2 the four weeks before. 0.0% of all new papers.

Explaining Hyperbolic Neural Networks via Geometry-Aware Relevance Propagation
Sep 28, 2026Ping Xiong, Shanglin Li, Yi Ding +2Neural Network Interpretability Layer-Wise Relevance Propagation
BIFOLD, Germany · Technische Universität Berlin, Germany · Nanyang Technological University, Singapore +3
ResLRP: The Role of Residual Cancellation in Attribution Instability in Vision Transformers
Sep 15, 2026Jim Berend, Reduan Achtibat, Daniel Schäffer +4Transformer Interpretability Residual Learning
Fraunhofer Heinrich Hertz Institute · DSC ScaDS.AI, Leipzig University · ICT Cluster, Singapore Institute of Technology, Singapore +3
HiLRP: Toward One Trustworthy Explanation for Vision Transformer: Conservation-Valid Attribution via Attention Primitives
Sep 1, 2026Sathiyamohan Nishankar, Pubudu Sanjeewani, Asanka Perera +1Transformer Interpretability Vision Transformer
Faculty of Engineering, University of Peradeniya, Sri Lanka · School of Computing Technologies, RMIT University, Melbourne, Australia · School of Engineering & Digital Technologies, University of Southern Queensland, Brisbane, Australia +1
Gradient-Skipping Relevance Propagation for Efficient Explainability of Vision Transformers
Jul 11, 2026Christopher Buratti, Michele Marchetti, Federica Parlapiano +3Transformer Interpretability Vision Transformer
DII, Polytechnic University of Marche · CHIMOMO, University of Modena and Reggio Emilia
IG-Lens: Exact Additive Probability Attribution Across Transformer Layers via Telescoping Integrated Gradients
Jun 29, 2026Duc Anh NguyenGradient-Based Attribution Integrated Gradients
Hanoi University of Science and Technology
Relevant Walk Search for Explaining Graph Neural Networks
May 22, 2026Ping Xiong, Thomas Schnake, Michael Gastegger +3Layer-Wise Relevance Propagation GNN Explainability
Technische Universität Berlin (TU Berlin) · BIFOLD – Berlin Institute for the Foundations of Learning and Data · Freie Universität Berlin (FU Berlin) +4
Efficient Higher-order Subgraph Attribution via Message Passing
May 21, 2026Ping Xiong, Thomas Schnake, Grégoire Montavon +2Feature Attribution Layer-Wise Relevance Propagation
Technische Universität Berlin (TU Berlin) · BIFOLD – Berlin Institute for the Foundations of Learning and Data · Department of Artificial Intelligence, Korea University, Seoul 136-713, Korea +2
From Clever Hans to Scientific Discovery: Interpreting EEG Foundational Transformers with LRP
May 12, 2026Justus Meyer zu Bexten, Nico Scherf, Bogdan Franczyk +1Transformer Interpretability EEG Decoding
Center for Scalable Data Analytics and Artificial Intelligence (ScaDS.AI) Dresden/Leipzig, Leipzig University · Neural Data Science and Statistical Computing, Max Planck Institute for Human Cognitive and Brain Sciences · Faculty of Economics, Leipzig University +1
Playing the network backward: A Game Theoretic Attribution Framework
May 7, 2026Jakob Paul Zimmermann, Jim Berend, Georg Loho +2Gradient-Based Attribution Transformer Interpretability
Fraunhofer Heinrich-Hertz-Institute, Berlin, Germany. · Technische Universität Berlin, Berlin, Germany. · Freie Universität Berlin, Berlin, Germany. +2
Mitigating Multimodal LLMs Hallucinations via Relevance Propagation at Inference Time
May 3, 2026Itai Allouche, Joseph KeshetMultimodal Grounding Inference-Time Optimization
Department of Electrical and Computer Engineering, Technion, Haifa, Israel
Evaluating Post-hoc Explanations of the Transformer-based Genome Language Model DNABERT-2
Apr 23, 2026Isabel Kurth, Paulo Yanez Sarmiento, Bernhard Y. RenardTransformer Interpretability Attention Mechanisms
Hasso Plattner Institute University of Potsdam
Contrastive Attribution in the Wild: An Interpretability Analysis of LLM Failures on Realistic Benchmarks
Apr 20, 2026Rongyuan Tan, Jue Zhang, Zhuozhao Li +3LLM Evaluation Feature Attribution
1Southern University of Science and Technology, China · *Work done during an internship at Microsoft. · 2Microsoft
Concept-based explanations of Segmentation and Detection models in Natural Disaster Management
Mar 24, 2026Samar Heydari, Jawher Said, Galip Ümit Yolcu +7Explainable Object Detection Efficient Semantic Segmentation
Department of Artificial Intelligence, Fraunhofer Heinrich Hertz Institute, Berlin, Germany · Department of Informatics, Aristotle University of Thessaloniki, Thessaloniki, Greece
Beyond Attention Heatmaps: How to Get Better Explanations for Multiple Instance Learning Models in Histopathology
Mar 9, 2026Mina Jamshidi Idaji, Julius Hense, Tom Neuhäuser +12Saliency Map Evaluation Multiple Instance Learning
Berlin Institute for the Foundations of Learning and Data, Berlin, Germany · Machine Learning Group, Technische Universität Berlin, Berlin, Germany · Department of Chemistry, Chemical Physics Theory Group, University of Toronto, Canada +9
Enhancing Physics-Informed Neural Networks with Domain-aware Fourier Features: Towards Improved Performance and Interpretable Results
Mar 3, 2026Alberto Miño Calero, Luis Salamanca, Konstantinos E. TatsisFourier Feature Embeddings Neural Network Interpretability
Department of Engineering Cybernetics NTNU, Trondheim, Norway · Swiss Data Science Center (SDSC) ETH Zürich, Zürich, Switzerland
X-REFINE: XAI-based RElevance input-Filtering and archItecture fiNe-tuning for channel Estimation
Feb 25, 2026Abdul Karim Gizzini, Yahia MedjahdiChannel Estimation Efficient Neural Network Inference
University of Paris-Est Créteil (UPEC), LISSI/TincNET, F-94400, Vitry-sur-Seine, France. · IMT Nord Europe, Institut Mines Télécom, Centre for Digital Systems, F-59653 Villeneuve d’Ascq, France.

Sep 28, 2026J/K move · Enter open · S save

Explaining Hyperbolic Neural Networks via Geometry-Aware Relevance Propagation

Ping Xiong, Shanglin Li, Yi Ding, Thomas Schnake+1

BIFOLD, Germany · Technische Universität Berlin, Germany · Nanyang Technological University, Singapore+3

Neural Network Interpretability Layer-Wise Relevance Propagation Hyperbolic Representation Learning

Open paper AskPDF arXiv

Hyperbolic neural networks introduce geometric operations that require explicit treatment in relevance propagation. Equivalent geometric realizations can produce different feature attributions, even when local relevance is conserved. We study this problem through Geometric Representation Invariance (GRI), a specialization of Implementation Invariance, and zero-curvature consistency, which requires identity relevance propagation when a geometric module approaches the identity. We propose LRP-radial-all for origin-centered radial modules, treating geometric scaling as modulation and assigning relevance entirely to the signal branch. The rule conserves relevance, is invariant to equivalent radial factorizations, and satisfies zero-curvature consistency, yielding GRI for a specified Poincaré-Lorentz logarithmic-map construction. In contrast, a conservative LRP-half baseline can violate both consistency criteria. Experiments on hyperbolic MNIST, sEEG, and CIFAR-10 classifiers assess attribution fidelity, qualitative explanations, and runtime. LRP-radial-all achieves competitive attribution fidelity across datasets with runtime comparable to Gradient

\times

Input and substantially lower than Integrated Gradients. These findings motivate geometry-aware propagation rules that distinguish relevance conservation from consistency across equivalent computations.