Research map / Learning theory and algorithms / Prediction and decision losses
Efficiently Computing Compact Formal Explanations
Min Wu, Xiaofu Li, Haoze Wu, Clark Barrett
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 |
|---|---|---|---|
| replace saliency order with bound propagation order (ablation) | VeriX+ binary search · GTSRB-CNN | explanation size: 569.0 -> 355.8 | Improved |
| use binary search traversal instead of sequential (ablation) | VeriX · MNIST-CNN | generation time: 398.9 -> 46.3 | Improved |
| add confidence ranking to CHECK (ablation) | VeriX · MNIST-CNN | generation time: 439.2 -> 398.9 | Improved |
| use QuickXplain instead of binary search (ablation) | VeriX+ · GTSRB-CNN | explanation size: 355.8 -> 355.3 size, 251.0 -> 330.1 s | Mixed |
| use bound propagation order instead of saliency (ablation) | VeriX+ binary search · MNIST-CNN | explanation size: 105.7 -> 100.8 | Improved |
| use bound propagation order instead of saliency (ablation) | VeriX+ binary search · CIFAR10-CNN | explanation size: 664.8 -> 553.5 | Improved |
| adversarial training of MNIST-FC | MNIST-FC · MNIST-FC original correct samples | explanation size: 177.20 -> 128.22 | Improved |
| adversarial training of MNIST-FC | MNIST-FC · MNIST-FC original correct samples | epsilon-robust rate: 4% -> 52.3% | Improved |
| use explanation size instead of max softmax for OOD detection | MNIST-CNN · CIFAR10 OOD | AUROC: 93% | Improved |
| use VeriX+ QuickXplain instead of Izza et al. Dichotomic search | VeriX+ · MNIST-FC | generation time: 82.3 -> 35.6 | Improved |
Closest papers by meaning
- Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanations (ICML 2025 · similarity 0.84)
- BoolXAI: Explainable AI Using Expressive Boolean Formulas (AAAI 2025 · similarity 0.84)
- Beyond Topological Self-Explainable GNNs: A Formal Explainability Perspective (ICML 2025 · similarity 0.84)
- GNNX-BENCH: Unravelling the Utility of Perturbation-based GNN Explainers through In-depth Benchmarking (ICLR 2024 · similarity 0.84)
- Even-if Explanations: Formal Foundations, Priorities and Complexity (AAAI 2025 · similarity 0.83)
- Efficient Reasoning via Constrained Optimization in Latent Space (NeurIPS 2026 · similarity 0.83)
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.
