mlsanityData to 14 Oct 2026

The Representational Limit of Scalar Interactions: An Interventional Decomposition

arXiv 2606.19410 ↗First posted 17 Jun 2026stat.ML cs.LGcited by 1

Not on the map. Why?

Abstract (arXiv)

Signed pairwise interaction scores fundamentally conflate uniqueness (U), redundancy (R), and synergy (S). We prove this on a minimal 3-way XOR structural causal model: faithful indices such as Shapley-Taylor return zero per pair, whereas projective indices such as Shapley Interaction spread the third-order effect into pair scalars that conflate the three mechanisms. We introduce Stochastic Hi-Fi, a post-hoc, retraining-free predictability decomposition that estimates per-feature U/R/S profiles by interventional masked inference. The estimator provides exact interventional semantics, finite-sample Monte Carlo bounds, strict variance reduction from coupled diamond sampling, and uniform finite-vocabulary convergence. Across tabular SCMs, Stochastic Hi-Fi recovers structure missed by scalar baselines (up to 411x larger interaction-magnitude recovery ratios). It also separates redundant and synergistic heads in the GPT-2 IOI circuit. On NIH ChestX-ray14, Stochastic Hi-Fi matches GradCAM on Pointing Game and improves substantially on Deletion AUC.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 32 of its 57 references that are in the map. First posted on arXiv 17 Jun 2026.

Builds on

32 of 57 references.

References · 32

  1. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  2. 2016Densely Connected Convolutional Networkscited by 6,720
  3. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  4. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  5. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  6. 2013Visualizing and Understanding Convolutional Networkscited by 2,290
  7. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,165
  8. 2017ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseasescited by 972
  9. 2018RISE: Randomized Input Sampling for Explanation of Black-box Modelscited by 838
  10. 2023Towards Automated Circuit Discovery for Mechanistic Interpretabilitycited by 724
  11. 2016Distribution-Free Predictive Inference For Regressioncited by 668
  12. 2017Interpretable Explanations of Black Boxes by Meaningful Perturbationcited by 612
20 more
  1. 2019The many Shapley values for model explanationcited by 344
  2. 2019Efficient and Accurate Estimation of Lipschitz Constants for Deep Neural Networkscited by 293
  3. 2022Quantus: An Explainable AI Toolkit for Responsible Evaluation of Neural Network Explanations and Beyondcited by 171
  4. 2019Feature relevance quantification in explainable AI: A causal problemcited by 167
  5. 2019The Shapley Taylor Interaction Indexcited by 143
  6. 2022OpenXAI: Towards a Transparent Evaluation of Model Explanationscited by 126
  7. 2020Causal Shapley Values: Exploiting Causal Knowledge to Explain Individual Predictions of Complex Modelscited by 107
  8. 2020Explaining Explanations: Axiomatic Feature Interactions for Deep Networkscited by 103
  9. 2020How does this interaction affect me? Interpretable attribution for feature interactionscited by 73
  10. 2022Faith-Shap: The Faithful Shapley Interaction Indexcited by 69
  11. 2024shapiq: Shapley Interactions for Machine Learningcited by 43
  12. 2024KernelSHAP-IQ: Weighted Least-Square Optimization for Shapley Interactionscited by 22
  13. 2019A Novel Approach to the Partial Information Decompositioncited by 17
  14. 2024DiffusionPID: Interpreting Diffusion via Partial Information Decompositioncited by 16
  15. 2024Explaining Graph Neural Networks via Structure-aware Interaction Indexcited by 11
  16. 2020Explaining Local, Global, And Higher-Order Interactions In Deep Learningcited by 8
  17. 2025SHAP Meets Tensor Networks: Provably Tractable Explanations with Parallelismcited by 8
  18. 2024Assessing high-order effects in feature importance via predictability decompositioncited by 2
  19. 2025STRIDE: Subset-Free Functional Decomposition for XAI in Tabular Settingscited by 1
  20. 2026H-Sets: Hessian-Guided Discovery of Set-Level Feature Interactions in Image Classifierscited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Conditional Co-Ablation: Recovering Self-Repair Backups in Transformer Circuitscited by 9