Perturbation-Based Feature Attribution

Ask about this topic

Overview

This field studies which input features influence a model’s prediction by changing or removing parts of an input and observing the effect. Methods include masking or occlusion, feature permutation, LIME- and SHAP-style sampling, counterfactual changes, and learned or adaptive perturbations across images, text, tabular data, graphs, and audio. Researchers assess explanations using deletion and insertion tests, stability under input changes, agreement between methods, and uncertainty estimates. A central challenge is that the choice of what to perturb and how to replace it can create unrealistic inputs and alter the explanation, motivating more informative, distribution-aware interventions and standardized evaluation.

Latest papers

CardsList

41 papers6 papers in the last four weeks, up 50% on the four weeks before. 0.0% of all new papers.

  1. No Single Metric Tells the Whole Story: A Multi-Dimensional Evaluation Framework for Uncertainty Attributions

    Mar 25, 2026Emily Schiller, Teodor Chiaburu, Marco Zullich +1Explainability EvaluationUncertainty Quantification

    XITASO GmbH IT & Software Solutions, Augsburg, Germany · The Artificial Intelligence and Cognitive Load Research Lab, University College Cork · Berliner Hochschule für Technik, Berlin, Germany +2

  2. Toward Faithful Segmentation Attribution via Benchmarking and Dual-Evidence Fusion

    Mar 23, 2026Abu Noman Md Sakib, OFM Riaz Rahman Aranya, Kevin Desai +1Feature AttributionExplainability Evaluation

    The University of Texas at San Antonio

  3. xplainfi: Feature Importance and Statistical Inference for Machine Learning in R

    Mar 16, 2026Lukas Burk, Fiona Katharina Ewald, Giuseppe Casalicchio +2Feature AttributionPerturbation-Based Feature Attribution

  4. Informative Perturbation Selection for Uncertainty-Aware Post-hoc Explanations

    Mar 16, 2026Sumedha Chugh, Ranjitha Prasad, Nazreen ShahPerturbation-Based Feature AttributionUncertainty Quantification

    Indraprastha Institute of Information Technology Delhi, India

  5. 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 EvaluationMultiple 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

  6. Measuring the Fragility of Trust: Devising Credibility Index via Explanation Stability (CIES) for Business Decision Support Systems

    Mar 5, 2026Alin-Gabriel Vaduva, Simona-Vasilica Oprea, Adela BaraExplainability EvaluationExplanation Stability

    Bucharest University of Economic Studies, Department of Economic Informatics and Cybernetics, no. 6 Piaţa Romană, Bucharest, Romania

  7. Missingness Bias Calibration in Feature Attribution Explanations

    Mar 5, 2026Shailesh Sridhar, Anton Xue, Eric WongFeature AttributionPerturbation-Based Feature Attribution

    University of Pennsylvania · University of Texas at Austin

  8. What You Read is What You Classify: Highlighting Attributions to Text and Text-Like Inputs

    Feb 27, 2026Daniel S. Berman, Brian Merritt, Stanley Ta +4Explainable Text ClassificationExplainable Artificial Intelligence

    Johns Hopkins Applied Physics Laboratory, 11100 Johns Hopkins Rd., Laurel, MD 20723, USA. · Current affiliation: Google DeepMind, 14–18 Handyside Street, London, N1C 4DN, United Kingdom

  9. One Permutation Is All You Need: Fast, Deterministic Feature Importance and Model Stress-Testing

    Dec 15, 2025Albert DoradorPerturbation-Based Feature AttributionAlgorithmic Auditing

  10. Frequency-Aware Model Parameter Explorer: A new attribution method for improving explainability

    Sep 25, 2025Ali Yavari, Alireza Mohamadi, Elham Beydaghi +2Feature AttributionPerturbation-Based Feature Attribution

    Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Vienna, Austria · Independent · Dept. of Computer Engineering, Sharif University of Technology, Tehran, Iran +1

  11. Explainable embeddings with Distance Explainer

    May 21, 2025Christiaan Meijer, E. G. Patrick BosFeature AttributionPerturbation-Based Feature Attribution

    Netherlands eScience Center, Science Park 402, 1098 XH Amsterdam, Netherlands · Center for Information Technology, University of Groningen, P.O. Box 11044, 9700 CA Groningen, Netherlands