Research map / Adversarial robustness and security / Black-box and transfer attacks
Robust and Stable Black Box Explanations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Optimize explanation fidelity against worst-case input perturbations using ROPE | Global post hoc explanations · Bail, Academic, and Health subgroup shifts | fidelity drop: 0% to 7% (ROPE); 13% to 23% (baselines) | Helped |
| Use multiple local robust explanations rather than one global explanation | ROPE logistic and ROPE dset · Bail, Academic, and Health | training fidelity | Helped |
| Use a single global robust model rather than multiple local robust explanations | ROPE logistic and ROPE dset · Original training distributions | fidelity | Hurt |
| Use ROPE to construct explanations from shifted data | Black-box explanation methods · Synthetic correlation shifts | fidelity drop: less than a 10% drop | Helped |
| Use ROPE rather than LIME, SHAP, or MUSE | Explanation generation under small data perturbations · Synthetic stability experiment | structural similarity: 18.21% to 21.08% more structurally similar | Helped |
| Use ROPE logistic multi instead of the baseline explanation methods | Logistic-regression black box · Bail | coefficient mismatch: 2.70 vs. 4.28 (SHAP); 4.37 (LIME) | Helped |
| Use ROPE dset multi instead of MUSE | Decision-set black box · Bail, single DS | rule match: 6.25 vs. 4.39 | Helped |
Closest papers by meaning
- Black-box Certification and Learning under Adversarial Perturbations (ICML 2020 · similarity 0.89)
- How to Robustify Black-Box ML Models? A Zeroth-Order Optimization Perspective (ICLR 2022 · similarity 0.89)
- Training for Stable Explanation for Free (NeurIPS 2024 · similarity 0.88)
- Promoting Counterfactual Robustness through Diversity (AAAI 2024 · similarity 0.88)
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (ICLR 2024 · similarity 0.87)
- Trust Regions for Explanations via Black-Box Probabilistic Certification (ICML 2024 · similarity 0.86)
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.