Research map / Probabilistic and scientific modeling / Variational and generative sampling
Synthesising Counterfactual Explanations via Label-Conditional Gaussian Mixture Variational Autoencoders
Junqi Jiang, Francesco Leofante, Antonio Rago, Francesca Toni
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 |
|---|---|---|---|
| use LAPACE (latent path CEs) instead of FACE-Interpolate | FACE-Interpolate · wine-RF actionability | portion satisfying constraints: 0.74 -> 1.00 | Improved |
| use LAPACE instead of FACE-Interpolate | FACE-Interpolate · heloc-NN actionability | portion satisfying constraints: 0.82 -> 1.00 | Improved |
| use LAPACE instead of FACE-Greedy | FACE-Greedy · wine-RF actionability | portion satisfying constraints: 0.88 -> 1.00 | Improved |
| use LAPACE instead of FACE-Greedy | FACE-Greedy · heloc-RF actionability | portion satisfying constraints: 0.96 -> 1.00 | Improved |
| use LAPACE instead of FACE-Greedy | FACE-Greedy · wine-NN and heloc-NN actionability | portion satisfying constraints: 1.00 -> 1.00 | No clear effect |
| gradient-based constraint correction (LAPACE-constrained) instead of naive feature edits | LAPACE-Naive · heloc and wine actionability | portion satisfying constraints: 1.00 -> 1.00 | No clear effect |
| move CE point from First to Last along the latent path | LAPACE · robustness to model changes | validity after retraining: 100% | Improved |
| use centroid CE (LAPACE-Last) instead of nearest-neighbour CEs | NNCE/RobXCE/DRCE · CE benchmark | plausibility (LOF): LOF below 1.0 | Improved |
| use LAPACE instead of DiCE/DRCE for diversity | DiCE and DRCE · mixed-type datasets | diversity: outperform on mixed-type, worse on continuous-only | Mixed |
| train classifier on L-GMVAE synthetic instead of real data | random forest · adult (RF classifier) | TSTR accuracy: 93.82% -> 81.13% | Hurt |
| train classifier on L-GMVAE synthetic instead of real data | random forest · heloc (RF classifier) | TSTR accuracy: 73.97% -> 71.07% | Hurt |
Closest papers by meaning
- CounterFlowNet: From Minimal Changes to Meaningful Counterfactual Explanations (NeurIPS 2026 · similarity 0.90)
- DiCoFlex: Model-Agnostic Diverse Counterfactuals with Flexible Control (NeurIPS 2025 · similarity 0.89)
- CF-OPT: Counterfactual Explanations for Structured Prediction (ICML 2024 · similarity 0.88)
- Robust Stochastic Graph Generator for Counterfactual Explanations (AAAI 2024 · similarity 0.88)
- Rethinking Visual Counterfactual Explanations Through Region Constraint (ICLR 2025 · similarity 0.88)
- LeapFactual: Reliable Visual Counterfactual Explanation Using Conditional Flow Matching (NeurIPS 2025 · similarity 0.88)
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.
