Research map / Adversarial robustness and security / Certified and empirical robustness
Lethal Dose Conjecture on Data Poisoning
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 |
|---|---|---|---|
| Use AutoAugment policies for the base learners | DPA on CIFAR-10 · CIFAR-10 | certified robustness: roughly doubled | Helped |
| Use the AutoAugment policy for SVHN for the base learner | DPA · GTSRB | certified robustness: more than tripled | Helped |
| Use AutoAugment for the base learner | Network-In-Network base learners · CIFAR-10 | test accuracy with limited data: similar accuracy as the baseline using only 1/2 of data | Helped |
| Use AutoAugment for the base learner | Network-In-Network base learner · GTSRB | test accuracy with limited data: similar accuracy as the baseline using only 1/4∼ 1/3 of data | Helped |
| Use radius nearest neighbors over HOG features as the base learner | DPA · CIFAR-10 | certified robustness | Helped |
Closest papers by meaning
- Improved Certified Defenses against Data Poisoning with (Deterministic) Finite Aggregation (ICML 2022 · similarity 0.92)
- Not All Poisons are Created Equal: Robust Training against Data Poisoning (ICML 2022 · similarity 0.90)
- Agnostic Learning under Targeted Poisoning: Optimal Rates and the Role of Randomness (NeurIPS 2025 · similarity 0.89)
- Like Oil and Water: Group Robustness Methods and Poisoning Defenses May Be at Odds (ICLR 2024 · similarity 0.88)
- Phantom Transfer: Data Poisoning can Survive Data-Level Defences (NeurIPS 2026 · similarity 0.88)
- Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples (NeurIPS 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.