Automatic Discovery of Visual Circuits
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914764850987008 |
|---|---|
| author | Rajaram, Achyuta Chowdhury, Neil Torralba, Antonio Andreas, Jacob Schwettmann, Sarah |
| author_facet | Rajaram, Achyuta Chowdhury, Neil Torralba, Antonio Andreas, Jacob Schwettmann, Sarah |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_14349 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Automatic Discovery of Visual Circuits Rajaram, Achyuta Chowdhury, Neil Torralba, Antonio Andreas, Jacob Schwettmann, Sarah Computer Vision and Pattern Recognition Artificial Intelligence 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. |
| title | Automatic Discovery of Visual Circuits |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2404.14349 |