mlsanityData to 14 Oct 2026

Automatic Discovery of Visual Circuits

arXiv 2404.14349 ↗First posted 22 Apr 2024cs.CV cs.AIcited by 13

Not on the map. Why?

Abstract (arXiv)

To date, most discoveries of network subcomponents that implement human-interpretable computations in deep vision models have involved close study of single units and large amounts of human labor. We explore scalable methods for extracting the subgraph of a vision model's computational graph that underlies recognition of a specific visual concept. We introduce a new method for identifying these subgraphs: specifying a visual concept using a few examples, and then tracing the interdependence of neuron activations across layers, or their functional connectivity. We find that our approach extracts circuits that causally affect model output, and that editing these circuits can defend large pretrained models from adversarial attacks.

Key facts. Cited by 13 papers in the map (data through 14 Oct 2026). Builds on 14 of its 24 references that are in the map. First posted on arXiv 22 Apr 2024.

Builds on

14 of 24 references.

References · 14

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  2. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  3. 2014Going Deeper with Convolutionscited by 4,241
  4. 2015Learning Deep Features for Discriminative Localizationcited by 2,436
  5. 2013Visualizing and Understanding Convolutional Networkscited by 2,290
  6. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,165
  7. 2017Network Dissection: Quantifying Interpretability of Deep Visual Representationscited by 708
  8. 2020Understanding the Role of Individual Units in a Deep Neural Networkcited by 255
  9. 2023Localizing Model Behavior with Path Patchingcited by 195
  10. 2014Object Detectors Emerge in Deep Scene CNNscited by 179
  11. 2022Natural Language Descriptions of Deep Visual Featurescited by 137
  12. 2021Low-Complexity Probing via Finding Subnetworkscited by 64
2 more
  1. 2023Multimodal Neurons in Pretrained Text-Only Transformerscited by 49
  2. 2022Assessing Gender Bias in Predictive Algorithms using eXplainable AIcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 7; 2026: 6.

13 papers.

Sort:

By area:

  1. 2025Internal Causal Mechanisms Robustly Predict Language Model Out-of-Distribution Behaviorscited by 11
  2. 2026Formal Mechanistic Interpretability: Automated Circuit Discovery with Provable Guaranteescited by 11
  3. 2025HyperDAS: Towards Automating Mechanistic Interpretability with Hypernetworkscited by 9
  4. 2025CoE: Chain-of-Explanation via Automatic Visual Concept Circuit Description and Polysemanticity Quantificationcited by 9
  5. 2025Granular Concept Circuits: Toward a Fine-Grained Circuit Discovery for Concept Representationscited by 5
  6. 2026Certified Circuits: Stability Guarantees for Mechanistic Circuitscited by 5
  7. 2025Decoding Vision Transformers: the Diffusion Steering Lenscited by 2
  8. 2025Reading Images Like Texts: Sequential Image Understanding in Vision-Language Modelscited by 2
  9. 2025Multimodal Function Vectors for Visual Relationscited by 1
  10. 2026Seeing Through Circuits: Faithful Mechanistic Interpretability for Vision Transformerscited by 1
  11. 2026Explaining the Explainer: Understanding the Inner Workings of Transformer-based Symbolic Regression Modelscited by 0
  12. 2026Inside-Out: Measuring Generalization in Vision Transformers Through Inner Workingscited by 0
1 more
  1. 2026From Interpretability Methods to Interpretable Modelscited by 0