mlsanityData to 7 Oct 2026

A Comparative Study of Faithfulness Metrics for Model Interpretability Methods

arXiv 2204.05514 ↗First posted 12 Apr 2022cs.CL cs.LGcited by 38

On the map: LLM reasoning → Prompting and chain-of-thought reasoning

Abstract (arXiv)

Interpretation methods to reveal the internal reasoning processes behind machine learning models have attracted increasing attention in recent years. To quantify the extent to which the identified interpretations truly reflect the intrinsic decision-making mechanisms, various faithfulness evaluation metrics have been proposed. However, we find that different faithfulness metrics show conflicting preferences when comparing different interpretations. Motivated by this observation, we aim to conduct a comprehensive and comparative study of the widely adopted faithfulness metrics. In particular, we introduce two assessment dimensions, namely diagnosticity and time complexity. Diagnosticity refers to the degree to which the faithfulness metric favours relatively faithful interpretations over randomly generated ones, and time complexity is measured by the average number of model forward passes. According to the experimental results, we find that sufficiency and comprehensiveness metrics have higher diagnosticity and lower time complexity than the other faithfulness metric

Key facts. Cited by 38 papers in the map (data through 7 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 22 of its 38 references that are in the map. First posted on arXiv 12 Apr 2022.

Builds on

22 of 38 references.

References · 22

  1. 2017Attention Is All You Needcited by 59,932
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  3. 2020Language Models are Few-Shot Learnerscited by 32,199
  4. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,266
  5. 2017A Unified Approach to Interpreting Model Predictionscited by 6,488
  6. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
  7. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  8. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
  9. 2015Character-level Convolutional Networks for Text Classificationcited by 2,293
  10. 2013Visualizing and Understanding Convolutional Networkscited by 2,278
  11. 2017Learning Important Features Through Propagating Activation Differencescited by 1,454
  12. 2017Towards A Rigorous Science of Interpretable Machine Learningcited by 1,400
10 more
  1. 2014Convolutional Neural Networks for Sentence Classificationcited by 1,124
  2. 2019Attention is not Explanationcited by 916
  3. 2019Attention is not not Explanationcited by 629
  4. 2020Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?cited by 590
  5. 2019ERASER: A Benchmark to Evaluate Rationalized NLP Modelscited by 523
  6. 2019Is Attention Interpretable?cited by 387
  7. 2018On the Robustness of Interpretability Methodscited by 310
  8. 2019One Explanation Does Not Fit All: A Toolkit and Taxonomy of AI Explainability Techniquescited by 164
  9. 2016"What is Relevant in a Text Document?": An Interpretable Machine Learning Approachcited by 53
  10. 2021Improving the Faithfulness of Attention-based Explanations with Task-specific Information for Text Classificationcited by 29

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 5; 2023: 8; 2024: 16; 2025: 6; 2026: 3.

38 papers.

Sort:

By area:

  1. 2023Explainability for Large Language Models: A Surveycited by 331
  2. 2023Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretabilitycited by 194
  3. 2022Towards Faithful Model Explanation in NLP: A Surveycited by 129
  4. 2023Faithfulness Tests for Natural Language Explanationscited by 107
  5. 2022MultiViz: Towards Visualizing and Understanding Multimodal Modelscited by 36
  6. 2024Verification and Refinement of Natural Language Explanations through LLM-Symbolic Theorem Provingcited by 35
  7. 2024Model Internals-based Answer Attribution for Trustworthy Retrieval-Augmented Generationcited by 25
  8. 2024Interpretable Concept Bottlenecks to Align Reinforcement Learning Agentscited by 23
  9. 2023Learning by Self-Explainingcited by 17
  10. 2023Incorporating Attribution Importance for Improving Faithfulness Metricscited by 14
  11. 2025Why and How LLMs Hallucinate: Connecting the Dots with Subsequence Associationscited by 14
  12. 2025A Causal Lens for Evaluating Faithfulness Metricscited by 12
26 more
  1. 2022The Solvability of Interpretability Evaluation Metricscited by 11
  2. 2023ALMANACS: A Simulatability Benchmark for Language Model Explainabilitycited by 11
  3. 2024Evaluating Readability and Faithfulness of Concept-based Explanationscited by 10
  4. 2023Goodhart's Law Applies to NLP's Explanation Benchmarkscited by 9
  5. 2023GInX-Eval: Towards In-Distribution Evaluation of Graph Neural Network Explanationscited by 9
  6. 2024Cross-Refine: Improving Natural Language Explanation Generation by Learning in Tandemcited by 8
  7. 2022An Attention Matrix for Every Decision: Faithfulness-based Arbitration Among Multiple Attention-Based Interpretations of Transformers in Text Classificationcited by 6
  8. 2024Beyond Label Attention: Transparency in Language Models for Automated Medical Coding via Dictionary Learningcited by 5
  9. 2024The Effect of Model Size on LLM Post-hoc Explainability via LIMEcited by 4
  10. 2024Interpretable end-to-end Neurosymbolic Reinforcement Learning agentscited by 4
  11. 2025Noiser: Bounded Input Perturbations for Attributing Large Language Modelscited by 4
  12. 2024Counterfactuals As a Means for Evaluating Faithfulness of Attribution Methods in Autoregressive Language Modelscited by 3
  13. 2024DILA: Dictionary Label Attention for Mechanistic Interpretability in High-dimensional Multi-label Medical Coding Predictioncited by 3
  14. 2024One Wave To Explain Them All: A Unifying Perspective On Feature Attributioncited by 3
  15. 2024SMOSE: Sparse Mixture of Shallow Experts for Interpretable Reinforcement Learning in Continuous Control Taskscited by 3
  16. 2025Attribution Explanations for Deep Neural Networks: A Theoretical Perspectivecited by 2
  17. 2022Truthful Meta-Explanations for Local Interpretability of Machine Learning Modelscited by 1
  18. 2024Review of Explainable Graph-Based Recommender Systemscited by 1
  19. 2025Beyond Patches: Mining Interpretable Part-Prototypes for Explainable AIcited by 1
  20. 2025CID: Measuring Feature Importance Through Counterfactual Distributionscited by 1
  21. 2026Toward Faithful Explanations in Acoustic Anomaly Detectioncited by 1
  22. 2024Towards a Research Community in Interpretable Reinforcement Learning: the InterpPol Workshopcited by 0
  23. 2024IMAFD: An Interpretable Multi-stage Approach to Flood Detection from time series Multispectral Datacited by 0
  24. 2024A Multi-Task Text Classification Pipeline with Natural Language Explanations: A User-Centric Evaluation in Sentiment Analysis and Offensive Language Identification in Greek Tweetscited by 0
  25. 2026A Practical Guide Towards Interpreting Time-Series Deep Clinical Predictive Models: A Reproducibility Studycited by 0
  26. 2026Tree-of-Evidence: Efficient "System 2" Search for Faithful Multimodal Groundingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Measuring Faithfulness Depends on How You Measure: Classifier Sensitivity in LLM Chain-of-Thought Evaluation
  2. 2026A Positive Case for Faithfulness: LLM Self-Explanations Help Predict Model Behavior
  3. 2023Benchmarking Faithfulness: Towards Accurate Natural Language Explanations in Vision-Language Tasks
  4. 2026Faithfulness Metrics Don't Measure Faithfulness: A Meta-Evaluation with Ground Truth
  5. 2024Are self-explanations from Large Language Models faithful?
  6. 2023On Measuring Faithfulness or Self-consistency of Natural Language Explanations
  7. 2025A Necessary Step toward Faithfulness: Measuring and Improving Consistency in Free-Text Explanations
  8. 2024The Probabilities Also Matter: A More Faithful Metric for Faithfulness of Free-Text Explanations in Large Language Models