Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Resilience in Sequential Prediction via Abstention
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 |
|---|---|---|---|
| Allow abstention on adversarial examples and measure error by mistakes plus abstentions on non-adversarial examples | Sequential prediction with adversarial injections · Adversarial injections, known marginal | misclassification error: d^2 \log T | Helped |
| Use higher-order shattering probabilities rather than only the disagreement region | Level-based learner · Adversarial injections, known marginal | abstention error: 6d | Helped |
| Use a reference function and modified leave-one-out disagreement to decide whether to predict or abstain | Structure-based learner · Adversarial injections, unknown marginal, VC dimension 1 | misclassification error: 2 \sqrt{T \log T} | Helped |
| Use a reference function and modified leave-one-out disagreement to decide whether to predict or abstain | Structure-based learner · Adversarial injections, unknown marginal, VC dimension 1 | abstention error: \sqrt{T\log T} | Helped |
| Use the smallest enclosing rectangle and the α-based rule for uncertain examples | Structure-based rectangle learner · Adversarial injections, unknown marginal, axis-aligned rectangles in \mathbb{R}^p | misclassification error: \sqrt{pT \log T} | Helped |
| Use the smallest enclosing rectangle and the α-based rule for uncertain examples | Structure-based rectangle learner · Adversarial injections, unknown marginal, axis-aligned rectangles in \mathbb{R}^p | abstention error: 2\sqrt{pT\log T} + 2p \log T | Helped |
| Abstain on the disagreement region and predict only where consistent hypotheses agree | Disagreement-based learner · Adversarial injections, thresholds in one dimension | misclassification error: 0 | Helped |
| Use interval structure to predict when an example lies between two examples with the same label, abstaining otherwise | Interval strategy · Adversarial injections, d-intervals in one dimension | misclassification error: \le 2d | Helped |
Closest papers by meaning
- Adversarial Robustness with Semi-Infinite Constrained Learning (NeurIPS 2021 · similarity 0.89)
- Posterior Mismatch Matters: Adversarial Training for Long-Tailed Robustness (ICML 2026 · similarity 0.88)
- Reliable Abstention under Adversarial Injections: Lower Bounds and New Upper Bounds (NeurIPS 2026 · similarity 0.88)
- Adversarially Robust Hypothesis Transfer Learning (ICML 2024 · similarity 0.87)
- Adversarial Robustness via Deformable Convolution with Stochasticity (ICML 2025 · similarity 0.87)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · 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.