XKDD (eXplaining Knowledge Discovery in Data Mining) is a workshop dedicated to eXplainable Artificial Intelligence (XAI) in Data Mining, focusing on methods that explain the behavior and outputs of complex Machine Learning models.
This edition focuses on the validation of explanation methods. While numerous XAI techniques have been proposed, the field still lacks a clear and widely accepted evaluation strategy, making it difficult to determine which explanations are reliable and informative.
For this reason, we invite contributions that examine how explanation methods can be evaluated and compared across models, tasks, and data distributions. We are particularly interested in contributions that analyze the assumptions and limitations of existing evaluation metrics and study their behavior across models, tasks, or data distributions. The interplay with ethical values is also an important topic, such as how reliably XAI methods reflect properties including faithfulness, stability, and usefulness.
Beyond quantitative evaluation, the workshop also addresses qualitative validation, including expert analysis and real-world use cases. In high-stakes settings involving sensitive data, numerical metrics alone may not reveal whether explanations are faithful, understandable, or aligned with ethical principles.
By focusing on validation as a core component of XAI, XKDD aims to strengthen the methodological foundations of explainability in Data Mining and Machine Learning.
Topics of interest include, but are not limited to:
Electronic submissions will be handled via CMT.
Papers must be written in English and formatted according to the Springer Lecture Notes in Computer Science (LNCS) guidelines, using the same format as the main conference.
The maximum length for both research papers and position papers is 16 pages, including references. Overlength papers will be rejected without review. Papers with smaller margins or font sizes than those specified in the author instructions and style files will also be treated as overlength.
We also accept full papers and 2 to 4 page abstracts, including references, for presentation only. These submissions may outline new emerging ideas and/or already published work, with the aim of stimulating discussion and collaboration among participants.
Authors who submit their work to XKDD 2026 commit to presenting their paper at the workshop if it is accepted. XKDD 2026 considers the author list submitted with the paper to be final. No additions or deletions may be made after submission, either during the review period or, in case of acceptance, at the final camera-ready stage.
When submitting a paper, authors should indicate in a note whether they wish to opt out of publication in the post-proceedings if the paper is accepted. All accepted full papers whose authors do not opt out will be published in the post-proceedings as part of the Lecture Notes in Computer Science series.
A condition for inclusion in the post-proceedings is that at least one co-author presents the paper at the workshop. Pre-proceedings will be made available online before the workshop.
All deadlines expire on 23:59 AoE.
Paper 325: Structure-Aware Explanation Evaluation in Autonomous Maritime Navigation
Arian Sabaghi; José Oramas
Paper 168: Measuring Monosemanticity in Sparse Autoencoders via Latent Activation Coherence
Katarzyna Filus; Sebastian Pokucinski
Paper 226: SAILS: Surrogate-based Analysis of Interactions via Local Effect Smooths
Timo Heiß; Julia Herbinger; Bernd Bischl; Giuseppe Casalicchio
Paper 301: Architecture Matters for Attribution Quality: A Quantitative Evaluation of XAI Methods Across Deep Vision Architectures in Artistic Style Classification
Maria Słomiany; OLGIERD UNOLD
Paper 138: HELM: A Hybrid Explainable Layered Model for Adaptive Post-Hoc Explanation Selection in Textual Toxicity Classification
Kossi Folly; Maria Malek
Paper 211: CEL: Comprehensive Counterfactual Explanations Library and Benchmark
Oleksii Furman; Łukasz Lenkiewicz; Marcel Musiałek; Maciej Zieba
Paper 243: A Probabilistic Consensus-Driven Approach for Robust Counterfactual Explanations
Marcin Kostrzewa; Maciej Zięba; Jerzy Stefanowski
Paper 255: Segment Attribution Tables: Segment-Level Representation and Aggregation of Saliency Maps
James Hinns; David Martens
Paper 370: Explaining Deep Learning Models Predicting Antimicrobial Resistance for Tuberculosis via Genomic Concepts
Paulo Yanez Sarmiento; Nadja Klein; Bernhard Renard
Paper 431: Validating Saliency Explanations as a Reliable Signal for YOLO-based Surveillance Detection
Claudio Giovannoni; Anna Monreale; Carlo Metta; Marco Natali
Paper 436: When Nothing Matters: Extending Linear Models to Improve Interpretability and Faithfulness
Martina Cinquini; Francesco Spinnato; Riccardo Guidotti
Paper 552: ReLAX-SC: Reinforcement Learning Agent eXplainer with Self-Correction
Alireza Heshmati; Panagiotis Papapetrou; Sindri Magnússon
Paper 115: Assessing Alignment and Stability of Feature Importance Explanations via Weight of Evidence
Eddie Conti; Claudio Daka; Álvaro Parafita; Antonio Luca Alfeo; Axel Brando; Mario G.C.A. Cimino
Paper 490: Interpretable Ensemble Anomaly Detection for Cybersecurity Time Series
Martina Cinquini; Alessio Cascione; Emanuele Sciancalepore; Riccardo Guidotti
Paper 479: Ontology-Guided Feature Engineering as an Explanation Regularizer
Filippos Gouidis; Theodore Patkos; Rommert Dekker; George Tzagkarakis
Paper 564: Counterfactual Transition Graphs: Evaluating Cross-Class Transition Quality
Syed Muhammad Hamza Zaidi; Szymon Bobek; Grzegorz Jacek Nalepa; Myra Spiliopoulou
Paper 182: Cross-Explainer Agreement as a Validation Diagnostic: A SHAP-LIME Study of Music Genre Classifiers
Zach Li; Yingtong Zhou; Yu Ding
Paper 209: Evaluation of Explainable AI for Time Series Classification: Frameworks and Open Challenges
Louis Peter; Nils Gumpfer; Jana Fischer; Christin Seifert; Jennifer Hannig
Peter Flach
Georgiana Ifrim
Paper 556: Predicting Compilation and Memory Failures from LLM Explanations
Ricardo Jarrin; Ritu Chaturvedi; Luiza Antonie
Paper 140: Explaining the Unsupervised: Human-in-the-Loop Supervision of Iterative Data Grouping
Gennady Andrienko
Paper 176: Robust and Interpretable Policy Learning for Manufacturing Process Parameters: From Policy Optimization to Explainable Recommendations
Yu-Hsin Hung; Bo-Ru Chen; Chia-Yen Lee
Paper 501: TRACE: Transparent Representation Auditing via Concept Extraction
Eleonora Poeta; Eliana Pastor; Tania Cerquitelli; Elena Baralis
Paper 454: A Paraconsistent Explanation Audit System for Conflicting Validation Evidence
Rafael Figueira Gonçalves; José Vinicius Ribeiro; Leonardo Arrighi; Sylvio Barbon Junior
Paper 442: Context-aware Spectral-Zone Explanation: A Case Study for Voice Pathology Detection
Sylvio Barbon Junior; José Vinicius Ribeiro; Rafael Figueira Gonçalves; Rodrigo Capobianco Guido
Paper 511: Rethinking Time Series Explanation Evaluation: A Temporal Coherence Metric Suite
Simon Ruttmann; Marcel Staiger; Felix Gerschner; Andreas Theissler
Paper 604: IMFACT: Counterfactual Explanations for Time Series via Intrinsic Mode Function Substitution
Udo Schlegel; Julian Rakuschek; Thomas Seidl; Andreas Holzinger; Tobias Schreck; Javier Del Ser
All 26 accepted papers.
The event will take place at the ECML-PKDD 2026 Conference, Room TBD.
Additional information about the location can be found at
the main conference web page: ECML-PKDD 2026
This workshop is partially supported by the European Community H2020 Program under research and innovation programme, grant agreement 834756 XAI, science and technology for the explanation of ai decision making.
This workshop is partially supported by TANGO. TANGO is a €7M EU-funded Horizon Europe project that aims to develop the theoretical foundations and the computational framework for synergistic human-machine decision making. The 4-year project will pave the way for the next generation of human-centric AI systems. TANGO.
This workshop is partially supported by the European Community NextGenerationEU programme under the funding schemes PNRR-PE-AI FAIR (Future Artificial Intelligence Research). FAIR.
This workshop has been partially supported by the Italian Project Fondo Italiano per la Scienza FIS00001966 ``MIMOSA''. MIMOSA.
The XKDD event was organised as part of the SoBigData.it project (Prot. IR0000013 - Call n. 3264 of 12/28/2021) initiatives aimed at training new users and communities in the usage of the research infrastructure (SoBigData.eu). “SoBigData.it receives funding from European Union – NextGenerationEU – National Recovery and Resilience Plan (Piano Nazionale di Ripresa e Resilienza, PNRR) – Project: “SoBigData.it – Strengthening the Italian RI for Social Mining and Big Data Analytics” – Prot. IR0000013 – Avviso n. 3264 del 28/12/2021.” SoBigData.it.
All inquires should be sent to
francesca.naretto@unipi.it
francesco.spinnato@unipi.it