Research map / Adversarial robustness and security / Certified and empirical robustness
Robust Adversarial Quantification via Conflict-Aware Evidential Deep Learning
Charmaine Barker, Daniel Bethell, Simos Gerasimou
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 conflict-aware evidence adjustment on top of metamorphic averaging (ablation) | EDL++ (Meta) · CIFAR10 -> SVHN | adversarial coverage (%): 2.35 -> 1.25 | Improved |
| add conflict-aware evidence adjustment on top of metamorphic averaging (ablation) | EDL++ (Meta) · CIFAR10 -> SVHN | OOD coverage (%): 6.59 -> 4.69 | Improved |
| add conflict-aware evidence adjustment on top of metamorphic averaging (ablation) | EDL++ (Meta) · MNIST -> FashionMNIST | adversarial coverage (%): 16.43 -> 15.51 | Improved |
| add conflict-aware evidence adjustment on top of metamorphic averaging (ablation) | EDL++ (Meta) · MNIST -> FashionMNIST | OOD coverage (%): 1.96 -> 2.00 | No clear effect |
| add conflict-aware adjustment to MC Dropout evidence (ablation) | EDL++ (MC) · MNIST -> FashionMNIST | adversarial coverage (%): 38.81 -> 21.62 | Improved |
| add conflict-aware adjustment to MC Dropout evidence (ablation) | EDL++ (MC) · MNIST -> KMNIST | OOD coverage (%): 11.67 -> 3.69 | Improved |
| use metamorphic transformations instead of MC Dropout for evidence diversity (ablation) | C-EDL · MNIST -> KMNIST | adversarial coverage (%): 12.40 -> 3.01 | Improved |
| use metamorphic transformations instead of MC Dropout for evidence diversity (ablation) | C-EDL · CIFAR10 -> SVHN | adversarial coverage (%): 9.39 -> 1.25 | Improved |
| use metamorphic transformations instead of MC Dropout (ablation) | C-EDL · MNIST -> FashionMNIST | OOD coverage (%): 1.77 -> 2.00 | Hurt |
| replace EDL with C-EDL (Meta) post hoc | EDL · MNIST -> FashionMNIST | adversarial coverage (%): 52.21 -> 15.51 | Improved |
| replace EDL with C-EDL (Meta) post hoc | EDL · CIFAR10 -> SVHN | adversarial coverage (%): 20.00 -> 1.25 | Improved |
| replace EDL with C-EDL (Meta) post hoc | EDL · MNIST -> FashionMNIST | ID coverage (%): 96.61 -> 94.18 | Hurt |
Closest papers by meaning
- Are Uncertainty Quantification Capabilities of Evidential Deep Learning a Mirage? (NeurIPS 2024 · similarity 0.91)
- Uncertainty Estimation by Flexible Evidential Deep Learning (NeurIPS 2025 · similarity 0.91)
- A Comprehensive Survey on Evidential Deep Learning and its Applications (TPAMI 2025 · similarity 0.91)
- Calibrating LLMs with Information-Theoretic Evidential Deep Learning (ICLR 2025 · similarity 0.90)
- Uncertainty Estimation by Density Aware Evidential Deep Learning (ICML 2024 · similarity 0.88)
- Detecting Misbehaviors of Large Vision-Language Models by Evidential Uncertainty Quantification (ICLR 2026 · 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.
