mlsanityData to 14 Oct 2026

FAME: Formal Abstract Minimal Explanation for Neural Networks

arXiv 2603.10661 ↗First posted 11 Mar 2026cs.AI cs.LGcited by 3

Not on the map. Why?

Abstract (arXiv)

We propose FAME (Formal Abstract Minimal Explanations), a new class of abductive explanations grounded in abstract interpretation. FAME is the first method to scale to large neural networks while reducing explanation size. Our main contribution is the design of dedicated perturbation domains that eliminate the need for traversal order. FAME progressively shrinks these domains and leverages LiRPA-based bounds to discard irrelevant features, ultimately converging to a formal abstract minimal explanation. To assess explanation quality, we introduce a procedure that measures the worst-case distance between an abstract minimal explanation and a true minimal explanation. This procedure combines adversarial attacks with an optional VERIX+ refinement step. We benchmark FAME against VERIX+ and demonstrate consistent gains in both explanation size and runtime on medium- to large-scale neural networks.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 34 of its 50 references that are in the map. First posted on arXiv 11 Mar 2026.

Builds on

34 of 50 references.

References · 34

  1. 2017Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networkscited by 744
  2. 2018Efficient Neural Network Robustness Certification with General Activation Functionscited by 445
  3. 2021Beta-CROWN: Efficient Bound Propagation with Per-neuron Split Constraints for Complete and Incomplete Neural Network Robustness Verificationcited by 228
  4. 2020Evaluating and Aggregating Feature-based Model Explanationscited by 188
  5. 2018Abduction-Based Explanations for Machine Learning Modelscited by 160
  6. 2018A Symbolic Approach to Explaining Bayesian Network Classifierscited by 131
  7. 2024Marabou 2.0: A Versatile Formal Analyzer of Neural Networkscited by 81
  8. 2020Explaining Naive Bayes and Other Linear Classifiers with Polynomial Time and Delaycited by 57
  9. 2021On Explaining Random Forests with SATcited by 57
  10. 2021On Guaranteed Optimal Robust Explanations for NLP Modelscited by 54
  11. 2020On Explaining Decision Treescited by 49
  12. 2022Logic-Based Explainability in Machine Learningcited by 48
22 more
  1. 2022Towards Formal XAI: Formally Approximate Minimal Explanations of Neural Networkscited by 42
  2. 2021SAT-Based Rigorous Explanations for Decision Listscited by 40
  3. 2021Explanations for Monotonic Classifierscited by 39
  4. 2022Don't Lie to Me! Robust and Efficient Explainability with Verified Perturbation Analysiscited by 38
  5. 2023The Fourth International Verification of Neural Networks Competition (VNN-COMP 2023): Summary and Resultscited by 38
  6. 2021On Efficiently Explaining Graph-Based Classifierscited by 37
  7. 2021A Review of Formal Methods applied to Machine Learningcited by 33
  8. 2021Trading Complexity for Sparsity in Random Forest Explanationscited by 32
  9. 2022VeriX: Towards Verified Explainability of Deep Neural Networkscited by 30
  10. 2020Achieving robustness in classification using optimal transport with hinge regularizationcited by 27
  11. 2022ASTERYX : A model-Agnostic SaT-basEd appRoach for sYmbolic and score-based eXplanationscited by 26
  12. 2022Xplique: A Deep Learning Explainability Toolboxcited by 24
  13. 2023Formally Explaining Neural Networks within Reactive Systemscited by 24
  14. 2022On the Computation of Necessary and Sufficient Explanationscited by 23
  15. 2024Distance-Restricted Explanations: Theoretical Underpinnings & Efficient Implementationcited by 21
  16. 2023From Robustness to Explainability and Back Againcited by 14
  17. 2025Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanationscited by 13
  18. 2023Disproving XAI Myths with Formal Methods -- Initial Resultscited by 12
  19. 2025Explain Yourself, Briefly! Self-Explaining Neural Networks with Concise Sufficient Reasonscited by 12
  20. 2016Propositional Abduction with Implicit Hitting Setscited by 6
  21. 2024On Using Certified Training towards Empirical Robustnesscited by 3
  22. 2024Surrogate Neural Networks Local Stability for Aircraft Predictive Maintenancecited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Incremental Neural Network Verification via Learned Conflictscited by 2
  2. 2026Beyond Explaining Predictions: Logic-Based Explanations for Confidence in Machine Learning Modelscited by 0
  3. 2026Beyond L2: Generalizing Abductive Latent Explanations to Diverse Prototype-Based Architecturescited by 0