Research map / Label-efficient and robust learning / Recognition and long-tail
DEPICT: Diffusion-Enabled Permutation Importance for Image Classification Tasks
Sarah Jabbour, Gregory Kondas, Ella Kazerooni, Michael Sjoding, David Fouhey, Jenna Wiens
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| permute concepts in text space and regenerate images with diffusion instead of using GradCAM/LIME IOU rankings | feature-importance explanation of image classifiers · Synthetic shapes | correlation with oracle weights: 0.73 -> 0.92 (vs LIME); 0.08 -> 0.92 (vs GradCAM) | Improved |
| DEPICT versus baselines on COCO primary feature models | DEPICT · COCO primary feature models | correlation with oracle: 0.15 -> 0.73 (vs LIME) | Improved |
| DEPICT versus baselines on COCO mixed feature models | DEPICT · COCO mixed feature models | correlation with oracle: 0.27 -> 0.35 (vs LIME), CIs overlap | Mixed |
| permute BMI/age/sex in text prompts of a chest X-ray diffusion model | MIMIC-CXR pneumonia classifiers · MIMIC-CXR | AUROC: sex: 1.00 -> 0.53; age: 0.85 -> 0.59; BMI: 0.97 -> 0.70 | Hurt |
| diffusion model conditioned on non-permutable text mentioning the permuted concept | DEPICT · MIMIC-CXR | independent permutation validity | Hurt |
| weaker agreement between DEPICT and bottleneck-permutation oracle | DEPICT · COCO | correlation with logistic regression weights: 0.98 (synthetic) -> 0.90 (primary) -> 0.49 (mixed) | Hurt |
Closest papers by meaning
- A Holistic Approach to Unifying Automatic Concept Extraction and Concept Importance Estimation (NeurIPS 2023 · similarity 0.87)
- Explaining in Diffusion: Explaining a Classifier with Diffusion Semantics (CVPR 2025 · similarity 0.86)
- Testing Semantic Importance via Betting (NeurIPS 2024 · similarity 0.85)
- Diffusion Classifiers Understand Compositionality, but Conditions Apply (NeurIPS 2025 · similarity 0.84)
- Probabilistic Conceptual Explainers: Trustworthy Conceptual Explanations for Vision Foundation Models (ICML 2024 · similarity 0.84)
- ConceptBed: Evaluating Concept Learning Abilities of Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
