Exploring SAIG Methods for an Objective Evaluation of XAI
Not on the map. Why?
Abstract (arXiv)
The evaluation of eXplainable Artificial Intelligence (XAI) methods is a rapidly growing field, characterized by a wide variety of approaches. This diversity highlights the complexity of the XAI evaluation, which, unlike traditional AI assessment, lacks a universally correct ground truth for the explanation, making objective evaluation challenging. One promising direction to address this issue involves the use of what we term Synthetic Artificial Intelligence Ground truth (SAIG) methods, which generate artificial ground truths to enable the direct evaluation of XAI techniques. This paper presents the first review and analysis of SAIG methods. We introduce a novel taxonomy to classify these approaches, identifying seven key features that distinguish different SAIG methods. Our comparative study reveals a concerning lack of consensus on the most effective XAI evaluation techniques, underscoring the need for further research and standardization in this area.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 54 of its 82 references that are in the map. First posted on arXiv 9 Feb 2026.
Builds on
54 of 82 references.
References · 54
- 2014Microsoft COCO: Common Objects in Contextcited by 17,498
- 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
- 2017A Unified Approach to Interpreting Model Predictionscited by 6,488
- 2017Axiomatic Attribution for Deep Networkscited by 3,350
- 2015Learning Deep Features for Discriminative Localizationcited by 2,430
- 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
- 2013Visualizing and Understanding Convolutional Networkscited by 2,278
- 2017Learning Important Features Through Propagating Activation Differencescited by 1,454
- 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,309
- 2017SmoothGrad: removing noise by adding noisecited by 1,097
- 2014Striving for Simplicity: The All Convolutional Netcited by 983
- 2024Are We on the Right Way for Evaluating Large Vision-Language Models?cited by 893
38 more
- 2020Quantifying Attention Flow in Transformerscited by 759
- 2018This Looks Like That: Deep Learning for Interpretable Image Recognitioncited by 731
- 2019Score-CAM: Score-Weighted Visual Explanations for Convolutional Neural Networkscited by 575
- 2015Evaluating the visualization of what a Deep Neural Network has learnedcited by 416
- 2016Top-down Neural Attention by Excitation Backpropcited by 318
- 2018On the Robustness of Interpretability Methodscited by 310
- 2019On the (In)fidelity and Sensitivity for Explanationscited by 309
- 2022From Anecdotal Evidence to Quantitative Evaluation Methods: A Systematic Review on Evaluating Explainable AIcited by 299
- 2017Right for the Right Reasons: Training Differentiable Models by Constraining their Explanationscited by 293
- 2019Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNetcited by 260
- 2024Vision language models are blind: Failing to translate detailed visual features into wordscited by 144
- 2021Benchmarking and Survey of Explanation Methods for Black Box Modelscited by 136
- 2020Ground Truth Evaluation of Neural Network Explanations with CLEVR-XAIcited by 125
- 2022OpenXAI: Towards a Transparent Evaluation of Model Explanationscited by 125
- 2021Do Feature Attribution Methods Correctly Attribute Features?cited by 98
- 2019Benchmarking Attribution Methods with Relative Feature Importancecited by 89
- 2017Learning how to explain neural networks: PatternNet and PatternAttributioncited by 84
- 2025Vision-Language Models Do Not Understand Negationcited by 69
- 2021"Will You Find These Shortcuts?" A Protocol for Evaluating the Faithfulness of Input Salience Methods for Text Classificationcited by 66
- 2021Convolutional Dynamic Alignment Networks for Interpretable Classificationscited by 54
- 2021Do Input Gradients Highlight Discriminative Features?cited by 52
- 2022Rethinking Stability for Attribution-based Explanationscited by 52
- 2023FunnyBirds: A Synthetic Vision Dataset for a Part-Based Analysis of Explainable AI Methodscited by 49
- 2018Local Explanation Methods for Deep Neural Networks Lack Sensitivity to Parameter Valuescited by 40
- 2023The Meta-Evaluation Problem in Explainable AI: Identifying Reliable Estimators with MetaQuantuscited by 37
- 2021Fast Axiomatic Attribution for Neural Networkscited by 35
- 2023B-cos Alignment for Inherently Interpretable CNNs and Vision Transformerscited by 31
- 2017Visual Explanation by Interpretation: Improving Visual Feedback Capabilities of Deep Neural Networkscited by 28
- 2022Investigating the fidelity of explainable artificial intelligence methods for applications of convolutional neural networks in geosciencecited by 28
- 2022Towards Better Understanding Attribution Methodscited by 28
- 2024On Behalf of the Stakeholders: Trends in NLP Model Interpretability in the Era of LLMscited by 24
- 2021Focus! Rating XAI Methods and Finding Biasescited by 20
- 2024Position: Explain to Question not to Justifycited by 20
- 2024Worse than Random? An Embarrassingly Simple Probing Evaluation of Large Multimodal Models in Medical VQAcited by 19
- 2021Sanity Simulations for Saliency Methodscited by 15
- 2023Assessing Fidelity in XAI post-hoc techniques: A Comparative Study with Ground Truth Explanations Datasetscited by 15
- 2020Quantifying Explainability of Saliency Methods in Deep Neural Networks with a Synthetic Datasetcited by 14
- 2024A comprehensive study on fidelity metrics for XAIcited by 14
50 most cited of 54 All 54 (JSON)
Built on by
Citing papers per year: 2026: 1.
1 paper.