Research map / Label-efficient and robust learning / Recognition and long-tail
Beyond Trivial Counterfactual Explanations With Diverse Valuable Explanations
Pau Rodríguez, Massimo Caccia, Alexandre Lacoste, Lee Zamparo, Issam Laradji, Laurent Charlin, David Vazquez
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 Fisher-information masking of influential latent dimensions (ablation) | DiVE · CelebA 'Young' classifier | non-trivial explanations (%): 43.51 -> 42.99 | No clear effect |
| add spectral clustering of the Fisher matrix (ablation) | DiVE_Fisher · CelebA 'Young' classifier | non-trivial explanations (%): 42.99 -> 51.07 | Improved |
| add spectral clustering to DiVE (Fisher masking) (ablation) | DiVE · CelebA 'Young' classifier | attributes changed per explanation: 4.81 -> 4.58 | Improved |
| use disentangled beta-TCVAE backbone for xGEM+ (ablation) | xGEM+ · CelebA 'Young' classifier | non-trivial explanations (%): 18.56 -> 34.62 | Improved |
| use beta-TCVAE backbone for xGEM+ (ablation) | xGEM+ · CelebA 'Young' classifier | attributes changed per explanation: 06.92 -> 06.70 | No clear effect |
| add perceptual reconstruction loss (ablation) | DiVE · CelebA Smiling FID (Overall) | FID: 55.9 -> 29.4 | Improved |
| add perceptual reconstruction loss (ablation) | DiVE · CelebA Young FID (Overall) | FID: 47.7 -> 33.8 | Improved |
| use DiVE instead of PE | counterfactual generation · CelebA Smiling FID (Overall) | FID: 35.8 -> 29.4 | Improved |
| use DiVE instead of PE | counterfactual generation · CelebA Young FID (Overall) | FID: 53.4 -> 33.8 | Improved |
| add Fisher information masking (ablation) | DiVE · CelebA human evaluation | real non-trivial counterfactuals (%): 38.65 -> 40.56 | Improved |
| add spectral clustering of Fisher matrix (ablation) | DiVE_Fisher · CelebA human evaluation | real non-trivial counterfactuals (%): 40.56 -> 41.90 | Improved |
| use DiVE instead of xGEM+ | bias detection · CelebA bias detection, f_biased, Smiling target | confounding metric (overall): 0.29 -> 0.22 | Mixed |
Closest papers by meaning
- Counterfactual Explanations for Misclassified Images: How Human and Machine Explanations Differ (Abstract Reprint) (AAAI 2024 · similarity 0.90)
- LeapFactual: Reliable Visual Counterfactual Explanation Using Conditional Flow Matching (NeurIPS 2025 · similarity 0.90)
- Promoting Counterfactual Robustness through Diversity (AAAI 2024 · similarity 0.89)
- Counterfactual Maps: What They Are and How to Find Them (NeurIPS 2026 · similarity 0.89)
- Rethinking Visual Counterfactual Explanations Through Region Constraint (ICLR 2025 · similarity 0.89)
- Model Reconstruction Using Counterfactual Explanations: A Perspective From Polytope Theory (NeurIPS 2024 · similarity 0.89)
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.
