mlsanityData to 14 Oct 2026

B-cos Networks: Alignment is All We Need for Interpretability

arXiv 2205.10268 ↗First posted 20 May 2022cs.CV stat.MLcited by 88

Not on the map. Why?

Abstract (arXiv)

We present a new direction for increasing the interpretability of deep neural networks (DNNs) by promoting weight-input alignment during training. For this, we propose to replace the linear transforms in DNNs by our B-cos transform. As we show, a sequence (network) of such transforms induces a single linear transform that faithfully summarises the full model computations. Moreover, the B-cos transform introduces alignment pressure on the weights during optimisation. As a result, those induced linear transforms become highly interpretable and align with task-relevant features. Importantly, the B-cos transform is designed to be compatible with existing architectures and we show that it can easily be integrated into common models such as VGGs, ResNets, InceptionNets, and DenseNets, whilst maintaining similar performance on ImageNet. The resulting explanations are of high visual quality and perform well under quantitative metrics for interpretability. Code available at https://www.github.com/moboehle/B-cos.

Key facts. Cited by 88 papers in the map (data through 14 Oct 2026). Builds on 36 of its 48 references that are in the map. First posted on arXiv 20 May 2022.

Builds on

36 of 48 references.

References · 36

  1. 2017Attention Is All You Needcited by 60,361
  2. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  3. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  4. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  5. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  6. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  7. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 7,005
  8. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  9. 2016Densely Connected Convolutional Networkscited by 6,720
  10. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  11. 2015Rethinking the Inception Architecture for Computer Visioncited by 5,475
  12. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,070
24 more
  1. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  2. 2015Learning Deep Features for Discriminative Localizationcited by 2,436
  3. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
  4. 2017Learning Important Features Through Propagating Activation Differencescited by 1,458
  5. 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,319
  6. 2018Sanity Checks for Saliency Mapscited by 1,156
  7. 2014Striving for Simplicity: The All Convolutional Netcited by 990
  8. 2018Robustness May Be at Odds with Accuracycited by 880
  9. 2018RISE: Randomized Input Sampling for Explanation of Black-box Modelscited by 838
  10. 2018This Looks Like That: Deep Learning for Interpretable Image Recognitioncited by 737
  11. 2014On the Number of Linear Regions of Deep Neural Networkscited by 381
  12. 2019Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNetcited by 260
  13. 2019Full-Gradient Representation for Neural Network Visualizationcited by 191
  14. 2013Maxout Networkscited by 187
  15. 2019Benchmarking Attribution Methods with Relative Feature Importancecited by 89
  16. 2017Cosine Normalization: Using Cosine Similarity Instead of Dot Product in Neural Networkscited by 59
  17. 2017Deep Hyperspherical Learningcited by 55
  18. 2021Convolutional Dynamic Alignment Networks for Interpretable Classificationscited by 54
  19. 2021Do Input Gradients Highlight Discriminative Features?cited by 52
  20. 2018Decoupled Networkscited by 45
  21. 2017Non-linear Convolution Filters for CNN-based Learningcited by 38
  22. 2019Bridging Adversarial Robustness and Gradient Interpretabilitycited by 22
  23. 2019Kervolutional Neural Networkscited by 20
  24. 2017Generalizing the Convolution Operator in Convolutional Neural Networkscited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 3; 2023: 26; 2024: 24; 2025: 22; 2026: 13.

88 papers.

Sort:

By area:

  1. 2024Discover-then-Name: Task-Agnostic Concept Bottlenecks via Automated Concept Discoverycited by 91
  2. 2022ProtoPFormer: Concentrating on Prototypical Parts in Vision Transformers for Interpretable Image Recognitioncited by 63
  3. 2022"Help Me Help the AI": Understanding How Explainability Can Support Human-AI Interactioncited by 63
  4. 2023FunnyBirds: A Synthetic Vision Dataset for a Part-Based Analysis of Explainable AI Methodscited by 50
  5. 2023Adversarial Counterfactual Visual Explanationscited by 47
  6. 2023Frozen Transformers in Language Models Are Effective Visual Encoder Layerscited by 39
  7. 2023Learning Support and Trivial Prototypes for Interpretable Image Classificationcited by 38
  8. 2023B-cos Alignment for Inherently Interpretable CNNs and Vision Transformerscited by 31
  9. 2023Text-to-Image Models for Counterfactual Explanations: a Black-Box Approachcited by 28
  10. 2023ICICLE: Interpretable Class Incremental Continual Learningcited by 24
  11. 2024Visual Concept Connectome (VCC): Open World Concept Discovery and their Interlayer Connections in Deep Modelscited by 22
  12. 2023PDiscoNet: Semantically consistent part discovery for fine-grained recognitioncited by 21
38 more
  1. 2025A Comprehensive Survey on Self-Interpretable Neural Networkscited by 18
  2. 2023Latent Diffusion Counterfactual Explanationscited by 17
  3. 2023Studying How to Efficiently and Effectively Guide Models with Explanationscited by 16
  4. 2024DCBM: Data-Efficient Visual Concept Bottleneck Modelscited by 16
  5. 2024LucidPPN: Unambiguous Prototypical Parts Network for User-centric Interpretable Computer Visioncited by 15
  6. 2024Good Teachers Explain: Explanation-Enhanced Knowledge Distillationcited by 14
  7. 2025Rashomon Sets for Prototypical-Part Networks: Editing Interpretable Models in Real-Timecited by 13
  8. 2025Leakage and Interpretability in Concept-Based Modelscited by 13
  9. 2024B-cosification: Transforming Deep Neural Networks to be Inherently Interpretablecited by 12
  10. 2023Mixture of Gaussian-distributed Prototypes with Generative Modelling for Interpretable and Trustworthy Image Recognitioncited by 11
  11. 2023Discriminative Feature Attributions: Bridging Post Hoc Explainability and Inherent Interpretabilitycited by 10
  12. 2023DeViL: Decoding Vision features into Languagecited by 10
  13. 2023Q-SENN: Quantized Self-Explaining Neural Networkscited by 9
  14. 2024Intrinsic User-Centric Interpretability through Global Mixture of Expertscited by 9
  15. 2023Interpretability-Aware Vision Transformercited by 8
  16. 2024Benchmarking the Attribution Quality of Vision Modelscited by 8
  17. 2024InfoDisent: Explainability of Image Classification Models by Information Disentanglementcited by 8
  18. 2025Label-free Concept Based Multiple Instance Learning for Gigapixel Histopathologycited by 8
  19. 2023Take 5: Interpretable Image Classification with a Handful of Featurescited by 7
  20. 2025EPIC: Explanation of Pretrained Image Classification Networks via Prototypecited by 7
  21. 2023Sum-of-Parts: Self-Attributing Neural Networks with End-to-End Learning of Feature Groupscited by 6
  22. 2024Can Implicit Bias Imply Adversarial Robustness?cited by 5
  23. 2024Comprehensive Attribution: Inherently Explainable Vision Model with Feature Detectorcited by 5
  24. 2024LineArt: A Knowledge-guided Training-free High-quality Appearance Transfer for Design Drawing with Diffusion Modelcited by 5
  25. 2023From Classification to Segmentation with Explainable AI: A Study on Crack Detection and Growth Monitoringcited by 4
  26. 2024What Sketch Explainability Really Means for Downstream Taskscited by 4
  27. 2025B-cos LM: Efficiently Transforming Pre-trained Language Models for Improved Explainabilitycited by 4
  28. 2025SIDE: Sparse Information Disentanglement for Explainable Artificial Intelligencecited by 4
  29. 2025Interpretable Decision-Making for End-to-End Autonomous Drivingcited by 4
  30. 2026DAVE: Distribution-aware Attribution via ViT Gradient Decompositioncited by 4
  31. 2026Towards Intrinsic Interpretability of Large Language Models:A Survey of Design Principles and Architecturescited by 4
  32. 2023Towards credible visual model interpretation with path attributioncited by 3
  33. 2023False Sense of Security: Leveraging XAI to Analyze the Reasoning and True Performance of Context-less DGA Classifierscited by 3
  34. 2023Towards interpretable-by-design deep learning algorithmscited by 3
  35. 2024Towards Explaining Hypercomplex Neural Networkscited by 3
  36. 2025VITAL: More Understandable Feature Visualization through Distribution Alignment and Relevant Information Flowcited by 3
  37. 2023Single Domain Generalization via Normalised Cross-correlation Based Convolutionscited by 2
  38. 2024B-Cos Aligned Transformers Learn Human-Interpretable Featurescited by 2

50 most cited of 88 All 88 (JSON)