mlsanityData to 14 Oct 2026

Network Dissection: Quantifying Interpretability of Deep Visual Representations

arXiv 1704.05796 ↗First posted 19 Apr 2017cs.CV cs.AIcited by 708

Not on the map. Why?

Cited by 43 in Mechanistic interpretability, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

We propose a general framework called Network Dissection for quantifying the interpretability of latent representations of CNNs by evaluating the alignment between individual hidden units and a set of semantic concepts. Given any CNN model, the proposed method draws on a broad data set of visual concepts to score the semantics of hidden units at each intermediate convolutional layer. The units with semantics are given labels across a range of objects, parts, scenes, textures, materials, and colors. We use the proposed method to test the hypothesis that interpretability of units is equivalent to random linear combinations of units, then we apply our method to compare the latent representations of various networks when trained to solve different supervised and self-supervised training tasks. We further analyze the effect of training iterations, compare networks trained with different initializations, examine the impact of network depth and width, and measure the effect of dropout and batch normalization on the interpretability of deep visual representations. We demonstrate that the proposed method can shed light on characteristics of CNN models and training methods that go beyond measurements of their discriminative power.

Key facts. Cited by 708 papers in the map (data through 14 Oct 2026). First posted on arXiv 19 Apr 2017.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 102; 2022: 124; 2023: 123; 2024: 120; 2025: 117; 2026: 122.

708 papers.

Sort:

By area:

  1. 2021Masked Feature Prediction for Self-Supervised Visual Pre-Trainingcited by 576
  2. 2023Finding Neurons in a Haystack: Case Studies with Sparse Probingcited by 334
  3. 2022Language in a Bottle: Language Model Guided Concept Bottlenecks for Interpretable Image Classificationcited by 312
  4. 2022From Anecdotal Evidence to Quantitative Evaluation Methods: A Systematic Review on Evaluating Explainable AIcited by 302
  5. 2022Post-hoc Concept Bottleneck Modelscited by 293
  6. 2023A Survey on Self-supervised Learning: Algorithms, Applications, and Future Trendscited by 207
  7. 2023Reconstructing the Mind's Eye: fMRI-to-Image with Contrastive Learning and Diffusion Priorscited by 160
  8. 2022CLIP-Dissect: Automatic Description of Neuron Representations in Deep Vision Networkscited by 147
  9. 2023Getting aligned on representational alignmentcited by 143
  10. 2022From Attribution Maps to Human-Understandable Explanations through Concept Relevance Propagationcited by 141
  11. 2022Natural Language Descriptions of Deep Visual Featurescited by 137
  12. 2023Similarity of Neural Network Models: A Survey of Functional and Representational Measurescited by 131
38 more
  1. 2022Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networkscited by 129
  2. 2021TS-CAM: Token Semantic Coupled Attention Map for Weakly Supervised Object Localizationcited by 123
  3. 2021Acquisition of Chess Knowledge in AlphaZerocited by 123
  4. 2024Interpreting CLIP with Sparse Linear Concept Embeddings (SpLiCE)cited by 119
  5. 2021Interpretable Deep Learning: Interpretation, Interpretability, Trustworthiness, and Beyondcited by 118
  6. 2022Long-Tailed Recognition via Weight Balancingcited by 107
  7. 2021HIVE: Evaluating the Human Interpretability of Visual Explanationscited by 105
  8. 2023Tracr: Compiled Transformers as a Laboratory for Interpretabilitycited by 104
  9. 2023Recent Advancements in End-to-End Autonomous Driving using Deep Learning: A Surveycited by 103
  10. 2021Guided Integrated Gradients: An Adaptive Path Method for Removing Noisecited by 102
  11. 2021Do Feature Attribution Methods Correctly Attribute Features?cited by 98
  12. 2023Concept-based Explainable Artificial Intelligence: A Surveycited by 98
  13. 2021An Interpretability Illusion for BERTcited by 97
  14. 2021A Comprehensive Taxonomy for Explainable Artificial Intelligence: A Systematic Survey of Surveys on Methods and Conceptscited by 91
  15. 2024Discover-then-Name: Task-Agnostic Concept Bottlenecks via Automated Concept Discoverycited by 91
  16. 2022Deep Unlearning via Randomized Conditionally Independent Hessianscited by 84
  17. 2021Transparency of Deep Neural Networks for Medical Image Analysis: A Review of Interpretability Methodscited by 83
  18. 2021Leveraging Sparse Linear Layers for Debuggable Deep Networkscited by 82
  19. 2022Distilling Model Failures as Directions in Latent Spacecited by 82
  20. 2021Explainability of deep vision-based autonomous driving systems: Review and challengescited by 80
  21. 2022Concept Bottleneck Model with Additional Unsupervised Conceptscited by 79
  22. 2024On the Origins of Linear Representations in Large Language Modelscited by 79
  23. 2023Successor Heads: Recurring, Interpretable Attention Heads In The Wildcited by 77
  24. 2023Copy Suppression: Comprehensively Understanding an Attention Headcited by 76
  25. 2025Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Modelscited by 73
  26. 2023Discover and Cure: Concept-aware Mitigation of Spurious Correlationcited by 71
  27. 2024Uncovering Safety Risks of Large Language Models through Concept Activation Vectorcited by 67
  28. 2022Finding Skill Neurons in Pre-trained Transformer-based Language Modelscited by 66
  29. 2021This Looks Like That... Does it? Shortcomings of Latent Space Prototype Interpretability in Deep Networkscited by 65
  30. 2021Editing a classifier by rewriting its prediction rulescited by 65
  31. 2022On the Symmetries of Deep Learning Models and their Internal Representationscited by 64
  32. 2022"Help Me Help the AI": Understanding How Explainability Can Support Human-AI Interactioncited by 63
  33. 2024A Multimodal Automated Interpretability Agentcited by 63
  34. 2023Explaining black box text modules in natural language with language modelscited by 62
  35. 2025Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignmentcited by 62
  36. 2022Post hoc Explanations may be Ineffective for Detecting Unknown Spurious Correlationcited by 59
  37. 2021Unsupervised Part Discovery from Contrastive Reconstructioncited by 57
  38. 2024Explainable AI for Safe and Trustworthy Autonomous Driving: A Systematic Reviewcited by 53

50 most cited of 708 All 708 (JSON)