Research map / Image generation / Autoregressive and latent generation
Designing Counterfactual Generators using Deep Model Inversion
Jayaraman J. Thiagarajan, Vivek Narayanaswamy, Deepta Rajan, Jia Liang, Akshay Chaudhari, Andreas Spanias
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 |
|---|---|---|---|
| Add DEP manifold consistency to LSO + INR (ablation) | LSO + INR counterfactual generator · CelebA (non-smiling to smiling) | CD: 0.26 $\pm$0.06 -> {0.08 $\pm$ 0.03} | Improved |
| Add DEP manifold consistency to LSO + INR (ablation) | LSO + INR counterfactual generator · ISIC 2018 (MEL to NEV) | CD: 0.25 $\pm$0.12 -> {0.11 $\pm$ 0.01} | Improved |
| Add DEP manifold consistency to LSO + INR (ablation) | LSO + INR counterfactual generator · CelebA (non-smiling to smiling) | Concentration: 0.28 $\pm$ 0.19 -> 0.18 $\pm$ 0.11 | Improved |
| Add DEP manifold consistency to LSO + INR (ablation) | LSO + INR counterfactual generator · ISIC 2018 (MEL to NEV) | Concentration: 0.29 $\pm$ 0.13 -> 0.22 $\pm$ 0.07 | Improved |
| Use INR instead of DIP as the image prior with DEP (ablation) | LSO + DIP + DEP · CelebA (non-smiling to smiling) | Concentration: 0.26 $\pm$0.15 -> 0.18 $\pm$ 0.11 | Improved |
| Use INR instead of DIP as the image prior with DEP (ablation) | LSO + DIP + DEP · ISIC 2018 (MEL to NEV) | Concentration: 0.25 $\pm$0.09 -> 0.22 $\pm$ 0.07 | Improved |
| Use DEP with INR instead of DIP (ablation) | LSO + DIP + DEP · CelebA (non-smiling to smiling) | CD: 0.15 $\pm$ 0.04 -> {0.08 $\pm$ 0.03} | Improved |
| Use DEP with INR instead of DIP (ablation) | LSO + DIP + DEP · ISIC 2018 (MEL to NEV) | CD: 0.17 $\pm$ 0.05 -> {0.11 $\pm$ 0.01} | Improved |
| Use DUQ instead of DEP for manifold consistency (ablation) | LSO + INR counterfactual generator · CelebA (non-smiling to smiling) | Meaningful pixel manipulations | No clear effect |
| Use ISO instead of LSO for semantic discrepancy (ablation) | Counterfactual generator with image priors · CelebA (non-smiling to smiling) | Meaningful target-class changes | Hurt |
| Use LSO with a stronger image prior instead of weak TV + L2 (ablation) | Counterfactual generator · CelebA (non-smiling to smiling) | Image quality | Improved |
| Adopt progressive optimization (ablation) | DISC counterfactual generator · CelebA (non-smiling to smiling) | Counterfactual quality | Improved |
Closest papers by meaning
- V-CECE: Visual Counterfactual Explanations via Conceptual Edits (NeurIPS 2025 · similarity 0.88)
- Model Reconstruction Using Counterfactual Explanations: A Perspective From Polytope Theory (NeurIPS 2024 · similarity 0.87)
- Counterfactual Explanations for Misclassified Images: How Human and Machine Explanations Differ (Abstract Reprint) (AAAI 2024 · similarity 0.87)
- Rethinking Visual Counterfactual Explanations Through Region Constraint (ICLR 2025 · similarity 0.87)
- Counterfactual Maps: What They Are and How to Find Them (NeurIPS 2026 · similarity 0.87)
- CF-OPT: Counterfactual Explanations for Structured Prediction (ICML 2024 · similarity 0.87)
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.
